2311skia,07编解码图像下

2023-11-26 23:15
文章标签 图像 07 编解码 2311skia

本文主要是介绍2311skia,07编解码图像下，希望对大家解决编程问题提供一定的参考价值，需要的开发者们随着小编来一起学习吧！

5,典型解码器的onDecode方法

可做为编解码图像库用法的参考.

(1)Jpeg

bool SkJPEGImageDecoder::onDecode(SkStream* stream, SkBitmap* bm, Mode mode) {
#ifdef TIME_DECODESkAutoTime atm("JPEG Decode");
#endifJPEGAutoClean autoClean;jpeg_decompress_struct  cinfo;skjpeg_source_mgr       srcManager(stream, this);skjpeg_error_mgr errorManager;set_error_mgr(&cinfo, &errorManager);//在调用`setjmp`前,要`实例化`所有对象,以便在错误时正确清理它们.if (setjmp(errorManager.fJmpBuf)) {return return_false(cinfo, *bm, "setjmp");}initialize_info(&cinfo, &srcManager);autoClean.set(&cinfo);int status = jpeg_read_header(&cinfo, true);if (status != JPEG_HEADER_OK) {return return_false(cinfo, *bm, "read_header");}/*试满足请求的`sampleSize`.因为`jpeg`(尽量)可更快,因此只需使用他们的`num/denomapi`来估算大小即可.*/int sampleSize = this->getSampleSize();set_dct_method(*this, &cinfo);SkASSERT(1 == cinfo.scale_num);cinfo.scale_denom = sampleSize;turn_off_visual_optimizations(&cinfo);const SkColorType colorType = this->getBitmapColorType(&cinfo);const SkAlphaType alphaType = kAlpha_8_SkColorType == colorType ?  kPremul_SkAlphaType : kOpaque_SkAlphaType;adjust_out_color_space_and_dither(&cinfo, colorType, *this);if (1 == sampleSize && SkImageDecoder::kDecodeBounds_Mode == mode) {//假设`A8`位图不透明,以避免`检查`每个单独的像素.它不太可能是不透明的,因为不透明的`A8`位图不会很有趣.否则,`jpeg`图像是不透明的.return bm->setInfo(SkImageInfo::Make(cinfo.image_width, cinfo.image_height, colorType, alphaType));}/*`image_width`和`image_height`是,在`jpeg_read_header()`之后可用的`原始尺寸`.要查看缩放的维度,必须调用`jpeg_start_decompress()`,然后读取`output_width`和`output_height.`*/if (!jpeg_start_decompress(&cinfo)) {/*如果在此失败了,如果他们只是想(子采样边界),可能仍有足够的信息来返回给调用者.如果`sampleSize`为`1`,则已可返回了.因此,只需检查是否为`kDecodeBounds_Mode`状态,及是否有有效的输出大小.这里失败的一个原因是没有足够的`流数据`来完成设置.然而,似乎很早就计算出来了`输出维度`,因此该特殊的检查有作用.*/if (SkImageDecoder::kDecodeBounds_Mode == mode && valid_output_dimensions(cinfo)) {SkScaledBitmapSampler smpl(cinfo.output_width, cinfo.output_height, recompute_sampleSize(sampleSize, cinfo));//假设`A8`位图不透明,以避免检查每个单独的像素.它不太可能是不透明的,因为不透明的`A8`位图不会很有趣.否则,`jpeg`图像是不透明的.return bm->setInfo(SkImageInfo::Make(smpl.scaledWidth(), smpl.scaledHeight(), colorType, alphaType));} else {return return_false(cinfo, *bm, "start_decompress");}}sampleSize = recompute_sampleSize(sampleSize, cinfo);
#ifdef SK_SUPPORT_LEGACY_IMAGEDECODER_CHOOSER//(如果有)是否应该允许选择器`选择`颜色类型if (!this->chooseFromOneChoice(colorType, cinfo.output_width, cinfo.output_height)) {return return_false(cinfo, *bm, "chooseFromOneChoice");}
#endifSkScaledBitmapSampler sampler(cinfo.output_width, cinfo.output_height, sampleSize);//假设`A8`位图不透明,以避免检查每个单独的像素.它不太可能是不透明的,因为不透明的`A8`位图不会很有趣.否则`,jpeg`图像是不透明的.bm->setInfo(SkImageInfo::Make(sampler.scaledWidth(), sampler.scaledHeight(), colorType, alphaType));if (SkImageDecoder::kDecodeBounds_Mode == mode) {return true;}if (!this->allocPixelRef(bm, NULL)) {return return_false(cinfo, *bm, "allocPixelRef");}SkAutoLockPixels alp(*bm);
#ifdef ANDROID_RGB/*如果可能,请短路`SkScaledBitmapSampler`,因为这会显著提高性能.*/if (sampleSize == 1 &&((kN32_SkColorType == colorType && cinfo.out_color_space == JCS_RGBA_8888) ||(kRGB_565_SkColorType == colorType && cinfo.out_color_space == JCS_RGB_565))){JSAMPLE* rowptr = (JSAMPLE*)bm->getPixels();INT32 const bpr =  bm->rowBytes();while (cinfo.output_scanline < cinfo.output_height) {int row_count = jpeg_read_scanlines(&cinfo, &rowptr, 1);if (0 == row_count) {//如果`row_count==0`,则没有得到扫描行,所以`提前返回`.返回部分图像.fill_below_level(cinfo.output_scanline, bm);cinfo.output_scanline = cinfo.output_height;break;  //跳到`jpeg_finish_decompress()`}if (this->shouldCancelDecode()) {return return_false(cinfo, *bm, "shouldCancelDecode");}rowptr += bpr;}jpeg_finish_decompress(&cinfo);return true;}
#endif//检查支持的格式SkScaledBitmapSampler::SrcConfig sc;int srcBytesPerPixel;if (!get_src_config(cinfo, &sc, &srcBytesPerPixel)) {return return_false(cinfo, *bm, "jpeg colorspace");}if (!sampler.begin(bm, sc, *this)) {return return_false(cinfo, *bm, "sampler.begin");}SkAutoMalloc srcStorage(cinfo.output_width * srcBytesPerPixel);uint8_t* srcRow = (uint8_t*)srcStorage.get();//可能跳过初始行`[sampler.srcY0]`if (!skip_src_rows(&cinfo, srcRow, sampler.srcY0())) {return return_false(cinfo, *bm, "skip rows");}//现在遍历扫描行,直到`y==bm->height()-1`for (int y = 0;; y++) {JSAMPLE* rowptr = (JSAMPLE*)srcRow;int row_count = jpeg_read_scanlines(&cinfo, &rowptr, 1);if (0 == row_count) {//如果`row_count==0`,则没有得到扫描行,所以提前返回.返回部分图像.fill_below_level(y, bm);cinfo.output_scanline = cinfo.output_height;break;  //跳到`jpeg_finish_decompress()`}if (this->shouldCancelDecode()) {return return_false(cinfo, *bm, "shouldCancelDecode");}if (JCS_CMYK == cinfo.out_color_space) {convert_CMYK_to_RGB(srcRow, cinfo.output_width);}sampler.next(srcRow);if (bm->height() - 1 == y) {//大功告成break;}if (!skip_src_rows(&cinfo, srcRow, sampler.srcDY() - 1)) {return return_false(cinfo, *bm, "skip rows");}}//正式跳过了其余部分,因此`libjpeg`不会抱怨if (!skip_src_rows(&cinfo, srcRow, cinfo.output_height - cinfo.output_scanline)) {return return_false(cinfo, *bm, "skip rows");}jpeg_finish_decompress(&cinfo);return true;
}

a,配置Jpeg解码选项(输出格式,采样率,idct类型,下采样率等等,在SkImageDecoder.h的注释中解释过一些)
b,在不设置下采样且目标格式为RGBA时,直接解码到目标Bitmap上(Jpeg是按YUV三分量分别压缩的,设置为Jpeg库解码时作YUV-RGBA的颜色转换).
c,需要下采样时,因为可简化jpeg库,因此重新计算一个下采样率,在目标位图上,下采样SkScaledBitmapSampler解码出来的图像.

(2)Png

bool SkPNGImageDecoder::onDecodeInit(SkStream* sk_stream, png_structp *png_ptrp, png_infop *info_ptrp) {/*使用期望错误处理器函数创建并初化`png_struct`.如果要用默认的`stderr`和`longjump`方法,可为最后三个参数提供`NULL`.还提供了编译器头文件版本,以便知道应用是否使用`兼容版本`的库编译.*/png_error_ptr user_warning_fn =(c_suppressPNGImageDecoderWarnings) ? (&do_nothing_warning_fn) : NULL;/*`NULL`表示保留为默认库行为.*//*`c_suppressPNGImageDecoderWarnings`默认值根据`SK_DEBUG`.*//*要使用`SK_DEBUG`二进制文件,禁止显示警告,请设置*环境变量`"skia_images_png_suppressDecoderWarnings"*`为`"true"`.在链接到`skia`的程序中:`*SK_CONF_SET("images.png.suppressDecoderWarnings",true);`*/png_structp png_ptr = png_create_read_struct(PNG_LIBPNG_VER_STRING, NULL, sk_error_fn, user_warning_fn);//`png_voidp user_error_ptr,user_error_fn,user_warning_fn);`if (png_ptr == NULL) {return false;}*png_ptrp = png_ptr;/*`分配/初化`图像信息的内存.*/png_infop info_ptr = png_create_info_struct(png_ptr);if (info_ptr == NULL) {png_destroy_read_struct(&png_ptr, png_infopp_NULL, png_infopp_NULL);return false;}*info_ptrp = info_ptr;/*如果使用`setjmp/longjmp`方法,请设置错误处理(这是使用`libpng`执行操作的普通方法).除非在前面`png_create_read_struct()`中设置了自己的错误处理器,否则为必需.*/if (setjmp(png_jmpbuf(png_ptr))) {png_destroy_read_struct(&png_ptr, &info_ptr, png_infopp_NULL);return false;}/*如果使用替换读函数,而不是在此处调用`*png_init_io()`,你可调用:*/png_set_read_fn(png_ptr, (void *)sk_stream, sk_read_fn);
#ifdef SK_BUILD_FOR_ANDROIDpng_set_seek_fn(png_ptr, sk_seek_fn);
#endif/*其中`user_io_ptr`是想可用来回调的结构*//*如果已阅读了一些签名*///`png_set_sig_bytes(png_ptr,0/*sig_read*/);`连接的偷看器,以便可见调用者可能感兴趣的用户块png_set_keep_unknown_chunks(png_ptr, PNG_HANDLE_CHUNK_ALWAYS, (png_byte*)"", 0);if (this->getPeeker()) {png_set_read_user_chunk_fn(png_ptr, (png_voidp)this->getPeeker(), sk_read_user_chunk);}/*调用`png_read_info()`,提供了第一个`IDAT`(图像数据块)之前的`PNG`文件中的所有信息.*/png_read_info(png_ptr, info_ptr);png_uint_32 origWidth, origHeight;int bitDepth, colorType;png_get_IHDR(png_ptr, info_ptr, &origWidth, &origHeight, &bitDepth, &colorType, int_p_NULL, int_p_NULL, int_p_NULL);/*告诉`libpng`去掉`16`位/颜色文件为8位/颜色*/if (bitDepth == 16) {png_set_strip_16(png_ptr);}/*将位深度为`1,2`和4的多个像素从单个字节提取进单独字节(适合调色板和灰度图像).*/if (bitDepth < 8) {png_set_packing(png_ptr);}/*将灰度图像从`1,2`或4位/像素扩展到完整的8位*/if (colorType == PNG_COLOR_TYPE_GRAY && bitDepth < 8) {png_set_expand_gray_1_2_4_to_8(png_ptr);}return true;
}
bool SkPNGImageDecoder::onDecode(SkStream* sk_stream, SkBitmap* decodedBitmap, Mode mode) {png_structp png_ptr;png_infop info_ptr;if (!onDecodeInit(sk_stream, &png_ptr, &info_ptr)) {return false;}PNGAutoClean autoClean(png_ptr, info_ptr);if (setjmp(png_jmpbuf(png_ptr))) {return false;}png_uint_32 origWidth, origHeight;int bitDepth, pngColorType, interlaceType;png_get_IHDR(png_ptr, info_ptr, &origWidth, &origHeight, &bitDepth, &pngColorType, &interlaceType, int_p_NULL, int_p_NULL);SkColorType         colorType;bool                hasAlpha = false;SkPMColor           theTranspColor = 0; //0告诉不要试匹配if (!this->getBitmapColorType(png_ptr, info_ptr, &colorType, &hasAlpha, &theTranspColor)) {return false;}SkAlphaType alphaType = this->getRequireUnpremultipliedColors() ?  kUnpremul_SkAlphaType : kPremul_SkAlphaType;const int sampleSize = this->getSampleSize();SkScaledBitmapSampler sampler(origWidth, origHeight, sampleSize);decodedBitmap->setInfo(SkImageInfo::Make(sampler.scaledWidth(), sampler.scaledHeight(), colorType, alphaType));if (SkImageDecoder::kDecodeBounds_Mode == mode) {return true;}//从此,关注的是`颜色表`和`跟踪`的像素,如果真看到`不透明`的像素,因为有时`PNG`设置其颜色类型为`|=PNG_COLOR_MASK_ALPHA`,但它的`所有像素`都是不透明的.小心,因为如果可标记位图不透明,绘画的速度会更快bool reallyHasAlpha = false;SkColorTable* colorTable = NULL;if (pngColorType == PNG_COLOR_TYPE_PALETTE) {decodePalette(png_ptr, info_ptr, &hasAlpha, &reallyHasAlpha, &colorTable);}SkAutoUnref aur(colorTable);if (!this->allocPixelRef(decodedBitmap, kIndex_8_SkColorType == colorType ? colorTable : NULL)) {return false;}SkAutoLockPixels alp(*decodedBitmap);/*打开隔行扫描处理.如果不使用`png_read_image()`,则为必须.要了解如何处理隔行扫描,见下面的`png_read_row()`方法:*/const int number_passes = (interlaceType != PNG_INTERLACE_NONE) ?  png_set_interlace_handling(png_ptr) : 1;/*可选调用`gamma`校正,并添加`背景`到调色板,并更新`信息结构`.如果想`libpng`为你更新调色板(即你在`上面`选择了`此转换`),则为必需.*/png_read_update_info(png_ptr, info_ptr);if ((kAlpha_8_SkColorType == colorType || kIndex_8_SkColorType == colorType) &&1 == sampleSize) {if (kAlpha_8_SkColorType == colorType) {//对`A8`位图,假设速度有一个`alpha`.位图可能是不透明的,但这是一个不太可能的用例,因为它不会很有趣.  reallyHasAlpha = true;//仅当原图为灰色时,才允许使用`A8`.SkASSERT(PNG_COLOR_TYPE_GRAY == pngColorType);}for (int i = 0; i < number_passes; i++) {for (png_uint_32 y = 0; y < origHeight; y++) {uint8_t* bmRow = decodedBitmap->getAddr8(0, y);png_read_rows(png_ptr, &bmRow, png_bytepp_NULL, 1);}}} else {SkScaledBitmapSampler::SrcConfig sc;int srcBytesPerPixel = 4;if (colorTable != NULL) {sc = SkScaledBitmapSampler::kIndex;srcBytesPerPixel = 1;} else if (kAlpha_8_SkColorType == colorType) {//仅当原图为灰色时,才允许使用`A8`.SkASSERT(PNG_COLOR_TYPE_GRAY == pngColorType);sc = SkScaledBitmapSampler::kGray;srcBytesPerPixel = 1;} else if (hasAlpha) {sc = SkScaledBitmapSampler::kRGBA;} else {sc = SkScaledBitmapSampler::kRGBX;}/*必须显式传递颜色表,因为要求升级`png`的调色板到直接模型,即使`decodedBitmap`没有,也可能有一个*/SkAutoLockColors ctLock(colorTable);if (!sampler.begin(decodedBitmap, sc, *this, ctLock.colors())) {return false;}const int height = decodedBitmap->height();if (number_passes > 1) {SkAutoMalloc storage(origWidth * origHeight * srcBytesPerPixel);uint8_t* base = (uint8_t*)storage.get();size_t rowBytes = origWidth * srcBytesPerPixel;for (int i = 0; i < number_passes; i++) {uint8_t* row = base;for (png_uint_32 y = 0; y < origHeight; y++) {uint8_t* bmRow = row;png_read_rows(png_ptr, &bmRow, png_bytepp_NULL, 1);row += rowBytes;}}//现在采样它.base += sampler.srcY0() * rowBytes;for (int y = 0; y < height; y++) {reallyHasAlpha |= sampler.next(base);base += sampler.srcDY() * rowBytes;}} else {SkAutoMalloc storage(origWidth * srcBytesPerPixel);uint8_t* srcRow = (uint8_t*)storage.get();skip_src_rows(png_ptr, srcRow, sampler.srcY0());for (int y = 0; y < height; y++) {uint8_t* tmp = srcRow;png_read_rows(png_ptr, &tmp, png_bytepp_NULL, 1);reallyHasAlpha |= sampler.next(srcRow);if (y < height - 1) {skip_src_rows(png_ptr, srcRow, sampler.srcDY() - 1);}}//(如果有)跳过其余行png_uint_32 read = (height - 1) * sampler.srcDY() +sampler.srcY0() + 1;SkASSERT(read <= origHeight);skip_src_rows(png_ptr, srcRow, origHeight - read);}}/*读取文件的其余部分,并在`info_ptr`中取其他块,必需*/png_read_end(png_ptr, info_ptr);if (0 != theTranspColor) {reallyHasAlpha |= substituteTranspColor(decodedBitmap, theTranspColor);}if (reallyHasAlpha && this->getRequireUnpremultipliedColors()) {switch (decodedBitmap->colorType()) {case kIndex_8_SkColorType://直降.case kARGB_4444_SkColorType://对这些颜色类型,选择不支持`unpremul`.return false;default: {//未完成解码.该颜色类型或支持`unpremul`,或不重要,因为它没有`alpha`(或只有`alpha`).这些括号可避免警告.//}}}if (!reallyHasAlpha) {decodedBitmap->setAlphaType(kOpaque_SkAlphaType);}return true;
}

Png格式图像可能包含透明度.Skia有个额外检查透明度是否全为255的,是则置不透明标志,毕竟在后续渲染不透明的图像时,消耗较少.

(3)Gif

bool SkGIFImageDecoder::onDecode(SkStream* sk_stream, SkBitmap* bm, Mode mode) {
#if GIFLIB_MAJOR < 5GifFileType* gif = DGifOpen(sk_stream, DecodeCallBackProc);
#elseGifFileType* gif = DGifOpen(sk_stream, DecodeCallBackProc, NULL);
#endifif (NULL == gif) {return error_return(*bm, "DGifOpen");}SkAutoTCallIProc<GifFileType, DGifCloseFile> acp(gif);SavedImage temp_save;temp_save.ExtensionBlocks=NULL;temp_save.ExtensionBlockCount=0;SkAutoTCallVProc<SavedImage, CheckFreeExtension> acp2(&temp_save);int width, height;GifRecordType recType;GifByteType *extData;
#if GIFLIB_MAJOR >= 5int extFunction;
#endifint transpIndex = -1;   //`-1`表示还没有它int fillIndex = gif->SBackGroundColor;do {if (DGifGetRecordType(gif, &recType) == GIF_ERROR) {return error_return(*bm, "DGifGetRecordType");}switch (recType) {case IMAGE_DESC_RECORD_TYPE: {if (DGifGetImageDesc(gif) == GIF_ERROR) {return error_return(*bm, "IMAGE_DESC_RECORD_TYPE");}if (gif->ImageCount < 1) {    //健全性检查return error_return(*bm, "ImageCount < 1");}width = gif->SWidth;height = gif->SHeight;SavedImage* image = &gif->SavedImages[gif->ImageCount-1];const GifImageDesc& desc = image->ImageDesc;int imageLeft = desc.Left;int imageTop = desc.Top;const int innerWidth = desc.Width;const int innerHeight = desc.Height;if (innerWidth <= 0 || innerHeight <= 0) {return error_return(*bm, "invalid dimensions");}//检查有效的描述符if (innerWidth > width) {gif_warning(*bm, "image too wide, expanding output to size");width = innerWidth;imageLeft = 0;} else if (imageLeft + innerWidth > width) {gif_warning(*bm, "shifting image left to fit");imageLeft = width - innerWidth;} else if (imageLeft < 0) {gif_warning(*bm, "shifting image right to fit");imageLeft = 0;}if (innerHeight > height) {gif_warning(*bm, "image too tall,  expanding output to size");height = innerHeight;imageTop = 0;} else if (imageTop + innerHeight > height) {gif_warning(*bm, "shifting image up to fit");imageTop = height - innerHeight;} else if (imageTop < 0) {gif_warning(*bm, "shifting image down to fit");imageTop = 0;}
#ifdef SK_SUPPORT_LEGACY_IMAGEDECODER_CHOOSER//`FIXME`:可让调用者选择图像或配置.if (!this->chooseFromOneChoice(kIndex_8_SkColorType, width, height)) {return error_return(*bm, "chooseFromOneChoice");}
#endifSkScaledBitmapSampler sampler(width, height, this->getSampleSize());bm->setInfo(SkImageInfo::Make(sampler.scaledWidth(), sampler.scaledHeight(), kIndex_8_SkColorType, kPremul_SkAlphaType));if (SkImageDecoder::kDecodeBounds_Mode == mode) {return true;}//现在解码`颜色表colortable`int colorCount = 0;{//在此处按域声明`colorPtr`.SkPMColor colorPtr[256]; //最坏时候的存储const ColorMapObject* cmap = find_colormap(gif);SkAlphaType alphaType = kOpaque_SkAlphaType;if (cmap != NULL) {SkASSERT(cmap->ColorCount == (1 << (cmap->BitsPerPixel)));colorCount = cmap->ColorCount;if (colorCount > 256) {colorCount = 256;  //`kIndex8`不能支持更多}for (int index = 0; index < colorCount; index++) {colorPtr[index] = SkPackARGB32(0xFF, cmap->Colors[index].Red, cmap->Colors[index].Green, cmap->Colors[index].Blue); }} else {//`find_colormap()`返回`NULL`.一些(罕见的,破碎的)`GIF`没有颜色表,所以强制使用颜色表.  gif_warning(*bm, "missing colormap");colorCount = 256;sk_memset32(colorPtr, SK_ColorWHITE, colorCount);}transpIndex = find_transpIndex(temp_save, colorCount);if (transpIndex >= 0) {colorPtr[transpIndex] = SK_ColorTRANSPARENT; //在透明的`SkPMColor`中`ram`alphaType = kPremul_SkAlphaType;fillIndex = transpIndex;} else if (fillIndex >= colorCount) {//`gif->SBackGroundColor`应小于`colorCount`.如果没有,请修复它.fillIndex = 0;  //}SkAutoTUnref<SkColorTable> ctable(SkNEW_ARGS(SkColorTable, (colorPtr, colorCount, alphaType)));if (!this->allocPixelRef(bm, ctable)) {return error_return(*bm, "allocPixelRef");}}//如果任一内部尺寸为`<=0`,则中止if (innerWidth <= 0 || innerHeight <= 0) {return error_return(*bm, "non-pos inner width/height");}SkAutoLockPixels alp(*bm);SkAutoMalloc storage(innerWidth);uint8_t* scanline = (uint8_t*) storage.get();//`GIF`有一个存储图像及由填充颜色填充的较大背景扫描行的`选项`.本例中,使用`较大位图`的子集采样.  SkBitmap subset;SkBitmap* workingBitmap;//只是`总边界`的一个子集吗if ((imageTop | imageLeft) > 0 ||innerWidth < width || innerHeight < height) {//填充背景.memset(bm->getPixels(), fillIndex, bm->getSize());//创建位图的子集.SkIRect subsetRect(SkIRect::MakeXYWH(imageLeft / sampler.srcDX(), imageTop / sampler.srcDY(), innerWidth / sampler.srcDX(), innerHeight / sampler.srcDY()));if (!bm->extractSubset(&subset, subsetRect)) {return error_return(*bm, "Extract failed.");}//更新采样器.现在,将仅采样子集.sampler = SkScaledBitmapSampler(innerWidth, innerHeight, this->getSampleSize());workingBitmap =  } else {workingBitmap = bm;}//已锁定`bm`了,但如果必须取一个子集,也必须锁定它,这样`getPixels()`才会指向它的像素.  SkAutoLockPixels alpWorking(*workingBitmap);if (!sampler.begin(workingBitmap, SkScaledBitmapSampler::kIndex, *this)) {return error_return(*bm, "Sampler failed to begin.");}//现在解码每条扫描行if (gif->Image.Interlace) {//遍历源数据的高度.采样器跳过不需要的行.GifInterlaceIter iter(innerHeight);for (int y = 0; y < innerHeight; y++) {if (DGifGetLine(gif, scanline, innerWidth) == GIF_ERROR) {gif_warning(*bm, "interlace DGifGetLine");memset(scanline, fillIndex, innerWidth);for (; y < innerHeight; y++) {sampler.sampleInterlaced(scanline, iter.currY());iter.next();}return true;}sampler.sampleInterlaced(scanline, iter.currY());iter.next();}} else {//简单,无隔行例const int outHeight = workingBitmap->height();skip_src_rows(gif, scanline, innerWidth, sampler.srcY0());for (int y = 0; y < outHeight; y++) {if (DGifGetLine(gif, scanline, innerWidth) == GIF_ERROR) {gif_warning(*bm, "DGifGetLine");memset(scanline, fillIndex, innerWidth);for (; y < outHeight; y++) {sampler.next(scanline);}return true;}//`Scanline`现在包含原始数据.试一试.sampler.next(scanline);if (y < outHeight - 1) {skip_src_rows(gif, scanline, innerWidth, sampler.srcDY() - 1);}}//(如果有)跳过其余行int read = (outHeight - 1) * sampler.srcDY() + sampler.srcY0() + 1;SkASSERT(read <= innerHeight);skip_src_rows(gif, scanline, innerWidth, innerHeight - read);}sanitize_indexed_bitmap(bm);return true;} break;case EXTENSION_RECORD_TYPE:
#if GIFLIB_MAJOR < 5if (DGifGetExtension(gif, &temp_save.Function, &extData) == GIF_ERROR) {
#elseif (DGifGetExtension(gif, &extFunction, &extData) == GIF_ERROR) {
#endifreturn error_return(*bm, "DGifGetExtension");}while (extData != NULL) {/*使用数据创建扩展块*/
#if GIFLIB_MAJOR < 5if (AddExtensionBlock(&temp_save, extData[0], &extData[1]) == GIF_ERROR) {
#elseif (GifAddExtensionBlock(&gif->ExtensionBlockCount, &gif->ExtensionBlocks, extFunction, extData[0], &extData[1]) == GIF_ERROR) {
#endifreturn error_return(*bm, "AddExtensionBlock");}if (DGifGetExtensionNext(gif, &extData) == GIF_ERROR) {return error_return(*bm, "DGifGetExtensionNext");}
#if GIFLIB_MAJOR < 5temp_save.Function = 0;
#endif}break;case TERMINATE_RECORD_TYPE:break;default:    /*`DGifGetRecordType`应抓它*/break;}} while (recType != TERMINATE_RECORD_TYPE);sanitize_indexed_bitmap(bm);return true;
}

因为适配不同版本的gif库的原因,这段代码写得略显杂乱,也存在一些八哥.

注意Skia对Gif动态图,只解最后一帧:

SavedImage* image = &gif->SavedImages[gif->ImageCount-1];

由SkGIFMovie类处理Gif动画,走不同的流程.应用需要调另外的类(详细见:frameworks/base/graphics/java/android/graphics/Movie.java).

(4)bmp

bool SkBMPImageDecoder::onDecode(SkStream* stream, SkBitmap* bm, Mode mode) {//首先读取整个流,以便传递`所有数据`给`BmpDecoderHelper`.分配保存数据的空间.SkAutoMalloc storage;//所有数据的字节长度.const size_t length = CopyStreamToStorage(&storage, stream);if (0 == length) {return 0;}const bool justBounds = SkImageDecoder::kDecodeBounds_Mode == mode;SkBmpDecoderCallback callback(justBounds);//现在按回调的`rgb()`数组`[r,g,b,r,g,b,...]`解码`BMP`{image_codec::BmpDecoderHelper helper;const int max_pixels = 16383*16383; //最大宽度*高度if (!helper.DecodeImage((const char*)storage.get(), length, max_pixels, &callback)) {return false;}}//不再需要它了,(在试分配位图的像素之前)而不是等待它的析构器,所以现在`释放`它storage.free();int width = callback.width();int height = callback.height();SkColorType colorType = this->getPrefColorType(k32Bit_SrcDepth, false);//只有在有`意义`时才接受`prefConfig`if (kARGB_4444_SkColorType != colorType && kRGB_565_SkColorType != colorType) {colorType = kN32_SkColorType;}SkScaledBitmapSampler sampler(width, height, getSampleSize());bm->setInfo(SkImageInfo::Make(sampler.scaledWidth(), sampler.scaledHeight(), colorType, kOpaque_SkAlphaType));if (justBounds) {return true;}if (!this->allocPixelRef(bm, NULL)) {return false;}SkAutoLockPixels alp(*bm);if (!sampler.begin(bm, SkScaledBitmapSampler::kRGB, *this)) {return false;}const int srcRowBytes = width * 3;const int dstHeight = sampler.scaledHeight();const uint8_t* srcRow = callback.rgb();srcRow += sampler.srcY0() * srcRowBytes;for (int y = 0; y < dstHeight; y++) {sampler.next(srcRow);srcRow += sampler.srcDY() * srcRowBytes;}return true;
}

这是最简单的一个格式,但用得较少,Skia里面的处理也比较随意.

a,把流中内容全部读到内存(CopyStreamToStorage)
b,解析流中的内容,用(image_codec::BmpDecoderHelper::DecodeImage)转化它为RGB格式
c,使用SkScaledBitmapSampler,来转换RGB为目标格式,同时下采样.

6,典型编码器的onEncode方法

(1)Jpeg

class SkJPEGImageEncoder : public SkImageEncoder {
protected:virtual bool onEncode(SkWStream* stream, const SkBitmap& bm, int quality) {
#ifdef TIME_ENCODESkAutoTime atm("JPEG Encode");
#endifSkAutoLockPixels alp(bm);if (NULL == bm.getPixels()) {return false;}jpeg_compress_struct    cinfo;skjpeg_error_mgr        sk_err;skjpeg_destination_mgr  sk_wstream(stream);//在设置调用`setjmp`前分配这些SkAutoMalloc    oneRow;SkAutoLockColors ctLocker;cinfo.err = jpeg_std_error(&sk_err);sk_err.error_exit = skjpeg_error_exit;if (setjmp(sk_err.fJmpBuf)) {return false;}//在`setjmp`或`markvolatile`之后保留.const WriteScanline writer = ChooseWriter(bm);if (NULL == writer) {return false;}jpeg_create_compress(&cinfo);cinfo.dest = &sk_wstream;cinfo.image_width = bm.width();cinfo.image_height = bm.height();cinfo.input_components = 3;
#ifdef WE_CONVERT_TO_YUVcinfo.in_color_space = JCS_YCbCr;
#elsecinfo.in_color_space = JCS_RGB;
#endifcinfo.input_gamma = 1;jpeg_set_defaults(&cinfo);jpeg_set_quality(&cinfo, quality, TRUE /*对基线`JPEG`值的限制*/);
#ifdef DCT_IFAST_SUPPORTEDcinfo.dct_method = JDCT_IFAST;
#endifjpeg_start_compress(&cinfo, TRUE);const int       width = bm.width();uint8_t*        oneRowP = (uint8_t*)oneRow.reset(width * 3);const SkPMColor* colors = ctLocker.lockColors(bm);const void*      srcRow = bm.getPixels();while (cinfo.next_scanline < cinfo.image_height) {JSAMPROW row_pointer[1];    /*`JSAMPLE`行指针*/writer(oneRowP, srcRow, width, colors);row_pointer[0] = oneRowP;(void) jpeg_write_scanlines(&cinfo, row_pointer, 1);srcRow = (const void*)((const char*)srcRow + bm.rowBytes());}jpeg_finish_compress(&cinfo);jpeg_destroy_compress(&cinfo);return true;}
};

Jpeg是有损压缩,传入的quality决定其量化参数表.
a,Skia默认先转换图片为YUV444格式,再编码(默认打开WE_CONVERT_TO_YUV宏,否则就是先转为RGB888格式,再传入Jpeg编码时转YUV).

b,默认使用JDCT_IFAST方法来傅立叶变换,很明显会造成一定的图片质量损失(即使quality设成100也存在,是计算精度的问题).

(2)Png

static transform_scanline_proc choose_proc(SkColorType ct, bool hasAlpha) {//如果是`kIndex8`,并不关心`alpha`上的搜索,因为只有打包`颜色表`关心该区别,而不是像素if (kIndex_8_SkColorType == ct) {hasAlpha = false;   //在`kIndex8`的表项中存储`false`}static const struct {SkColorType             fColorType;bool                    fHasAlpha;transform_scanline_proc fProc;} gMap[] = {{ kRGB_565_SkColorType,     false,  transform_scanline_565 },{ kN32_SkColorType,         false,  transform_scanline_888 },{ kN32_SkColorType,         true,   transform_scanline_8888 },{ kARGB_4444_SkColorType,   false,  transform_scanline_444 },{ kARGB_4444_SkColorType,   true,   transform_scanline_4444 },{ kIndex_8_SkColorType,     false,  transform_scanline_memcpy },};for (int i = SK_ARRAY_COUNT(gMap) - 1; i >= 0; --i) {if (gMap[i].fColorType == ct && gMap[i].fHasAlpha == hasAlpha) {return gMap[i].fProc;}}sk_throw();return NULL;
}
bool SkPNGImageEncoder::onEncode(SkWStream* stream, const SkBitmap& bitmap, int /*质量*/) {SkColorType ct = bitmap.colorType();const bool hasAlpha = !bitmap.isOpaque();int colorType = PNG_COLOR_MASK_COLOR;int bitDepth = 8;   //颜色的默认值png_color_8 sig_bit;switch (ct) {case kIndex_8_SkColorType:colorType |= PNG_COLOR_MASK_PALETTE;//直降至`ARGB_8888`例case kN32_SkColorType:sig_bit.red = 8;sig_bit.green = 8;sig_bit.blue = 8;sig_bit.alpha = 8;break;case kARGB_4444_SkColorType:sig_bit.red = 4;sig_bit.green = 4;sig_bit.blue = 4;sig_bit.alpha = 4;break;case kRGB_565_SkColorType:sig_bit.red = 5;sig_bit.green = 6;sig_bit.blue = 5;sig_bit.alpha = 0;break;default:return false;}if (hasAlpha) {//如果是调色板,则即使`ctable`有`alpha`,也不要指定`alpha`if (!(colorType & PNG_COLOR_MASK_PALETTE)) {colorType |= PNG_COLOR_MASK_ALPHA;}} else {sig_bit.alpha = 0;}SkAutoLockPixels alp(bitmap);//`readyToDraw`检查像素(如果需要,颜色表)if (!bitmap.readyToDraw()) {return false;}//必须在锁定像素后才这样SkColorTable* ctable = bitmap.getColorTable();if (NULL != ctable) {if (ctable->count() == 0) {return false;}//检查是否可按小于8位存储bitDepth = computeBitDepth(ctable->count());}return doEncode(stream, bitmap, hasAlpha, colorType, bitDepth, ct, sig_bit);
}
bool SkPNGImageEncoder::doEncode(SkWStream* stream, const SkBitmap& bitmap, const bool& hasAlpha, int colorType, int bitDepth, SkColorType ct, png_color_8& sig_bit) {png_structp png_ptr;png_infop info_ptr;png_ptr = png_create_write_struct(PNG_LIBPNG_VER_STRING, NULL, sk_error_fn, NULL);if (NULL == png_ptr) {return false;}info_ptr = png_create_info_struct(png_ptr);if (NULL == info_ptr) {png_destroy_write_struct(&png_ptr,  png_infopp_NULL);return false;}/*设置错误处理.如果未在`png_create_write_struct()`调用中提供自己的错误处理函数,则为必填项.*/if (setjmp(png_jmpbuf(png_ptr))) {png_destroy_write_struct(&png_ptr, &info_ptr);return false;}png_set_write_fn(png_ptr, (void*)stream, sk_write_fn, png_flush_ptr_NULL);/*在此处设置图像信息.宽度和高度最大为`2^31`,`bit_depth`是`1,2,4,8`或`16`之一,但有效值还根据所选的`color_type`.`color_type`是`PNG_COLOR_TYPE_GRAY,PNG_COLOR_TYPE_GRAY_ALPHA,PNG_COLOR_TYPE_PALETTE,PNG_COLOR_TYPE_RGB`,或`PNG_COLOR_TYPE_RGB_ALPHA`之一.隔行扫描是`PNG_INTERLACE_NONE`或`PNG_INTERLACE_ADAM7,compression_type`和`filter_type`当前必须为`PNG_COMPRESSION_TYPE_BASE`和`PNG_FILTER_TYPE_BASE`.必填*/png_set_IHDR(png_ptr, info_ptr, bitmap.width(), bitmap.height(), bitDepth, colorType, PNG_INTERLACE_NONE, PNG_COMPRESSION_TYPE_BASE, PNG_FILTER_TYPE_BASE);//如果需要,设置`colortable/trans`数组png_color paletteColors[256];png_byte trans[256];if (kIndex_8_SkColorType == ct) {SkColorTable* ct = bitmap.getColorTable();int numTrans = pack_palette(ct, paletteColors, trans, hasAlpha);png_set_PLTE(png_ptr, info_ptr, paletteColors, ct->count());if (numTrans > 0) {png_set_tRNS(png_ptr, info_ptr, trans, numTrans, NULL);}}png_set_sBIT(png_ptr, info_ptr, &sig_bit);png_write_info(png_ptr, info_ptr);const char* srcImage = (const char*)bitmap.getPixels();SkAutoSMalloc<1024> rowStorage(bitmap.width() << 2);char* storage = (char*)rowStorage.get();transform_scanline_proc proc = choose_proc(ct, hasAlpha);for (int y = 0; y < bitmap.height(); y++) {png_bytep row_ptr = (png_bytep)storage;proc(srcImage, bitmap.width(), storage);png_write_rows(png_ptr, &row_ptr, 1);srcImage += bitmap.rowBytes();}png_write_end(png_ptr, info_ptr);/*写入后清理,并释放分配的所有内存*/png_destroy_write_struct(&png_ptr, &info_ptr);return true;
}

a,统一把图像格式转为RGBA8888,再按png编码,另外注明是否包含透明度信息.
b,对原先带透明的图像格式(RGBA8888,RGBA4444),做反alpha预乘,也即每个像素值除以其alpha值,自然,会由乘法替换除法.

这篇关于2311skia,07编解码图像下的文章就介绍到这儿，希望我们推荐的文章对编程师们有所帮助！


http://www.chinasem.cn/article/426469。

相关文章

基于WinForm+Halcon实现图像缩放与交互功能

基于WinForm+Halcon实现图像缩放与交互功能

《基于WinForm+Halcon实现图像缩放与交互功能》本文主要讲述在WinForm中结合Halcon实现图像缩放、平移及实时显示灰度值等交互功能,包括初始化窗口的不同方式,以及通过特定事件添加相应... 目录前言初始化窗口添加图像缩放功能添加图像平移功能添加实时显示灰度值功能示例代码总结最后前言本文将
阅读更多...
基于人工智能的图像分类系统

基于人工智能的图像分类系统

目录 引言项目背景环境准备 硬件要求软件安装与配置系统设计 系统架构关键技术代码示例 数据预处理模型训练模型预测应用场景结论 1. 引言 图像分类是计算机视觉中的一个重要任务，目标是自动识别图像中的对象类别。通过卷积神经网络（CNN）等深度学习技术，我们可以构建高效的图像分类系统，广泛应用于自动驾驶、医疗影像诊断、监控分析等领域。本文将介绍如何构建一个基于人工智能的图像分类系统，包括环境
阅读更多...
Verybot之OpenCV应用一：安装与图像采集测试

Verybot之OpenCV应用一：安装与图像采集测试

在Verybot上安装OpenCV是很简单的，只需要执行：         sudo apt-get update         sudo apt-get install libopencv-dev         sudo apt-get install python-opencv         下面就对安装好的OpenCV进行一下测试，编写一个通过USB摄像头采
阅读更多...
07 v-if和v-show使用和区别

07 v-if和v-show使用和区别

划重点： v-ifv-show 小葱拌豆腐 <!DOCTYPE html><html lang="en"><head><meta charset="UTF-8"><meta name="viewport" content="width=device-width, initial-scale=1.0"><meta http-equiv="X-UA-Compatible" content="
阅读更多...
【python计算机视觉编程——7.图像搜索】

【python计算机视觉编程——7.图像搜索】

python计算机视觉编程——7.图像搜索 7.图像搜索7.1 基于内容的图像检索（CBIR）从文本挖掘中获取灵感——矢量空间模型（BOW表示模型）7.2 视觉单词**思想****特征提取**： 创建词汇7.3 图像索引7.3.1 建立数据库7.3.2 添加图像 7.4 在数据库中搜索图像7.4.1 利用索引获取获选图像7.4.2 用一幅图像进行查询7.4.3 确定对比基准并绘制结果 7.
阅读更多...
【python计算机视觉编程——8.图像内容分类】

【python计算机视觉编程——8.图像内容分类】

python计算机视觉编程——8.图像内容分类 8.图像内容分类8.1 K邻近分类法（KNN）8.1.1 一个简单的二维示例8.1.2 用稠密SIFT作为图像特征8.1.3 图像分类:手势识别 8.2贝叶斯分类器用PCA降维 8.3 支持向量机8.3.2 再论手势识别 8.4 光学字符识别8.4.2 选取特征8.4.3 多类支持向量机8.4.4 提取单元格并识别字符8.4.5 图像校正
阅读更多...
HalconDotNet中的图像特征与提取详解

HalconDotNet中的图像特征与提取详解

文章目录 简介一、边缘特征提取二、角点特征提取三、区域特征提取四、纹理特征提取五、形状特征提取 简介   图像特征提取是图像处理中的一个重要步骤，用于从图像中提取有意义的特征，以便进行进一步的分析和处理。HalconDotNet提供了多种图像特征提取方法，每种方法都有其特定的应用场景和优缺点。 一、边缘特征提取   边缘特征提取是图像处理中最基本的特征提取方法之一，通过检
阅读更多...
超越IP-Adapter！阿里提出UniPortrait，可通过文本定制生成高保真的单人或多人图像。

超越IP-Adapter！阿里提出UniPortrait，可通过文本定制生成高保真的单人或多人图像。

阿里提出UniPortrait，能根据用户提供的文本描述，快速生成既忠实于原图又能灵活调整的个性化人像，用户甚至可以通过简单的句子来描述多个不同的人物，而不需要一一指定每个人的位置。这种设计大大简化了用户的操作，提升了个性化生成的效率和效果。 UniPortrait以统一的方式定制单 ID 和多 ID 图像，提供高保真身份保存、广泛的面部可编辑性、自由格式的文本描述，并且无需预先确定的布局。
阅读更多...
Winfrom中解决图像、文字模糊的方法

Winfrom中解决图像、文字模糊的方法

1.添加清单 2.将清单中的下面内容取消注释
阅读更多...
使用亚马逊Bedrock的Stable Diffusion XL模型实现文本到图像生成：探索AI的无限创意

使用亚马逊Bedrock的Stable Diffusion XL模型实现文本到图像生成：探索AI的无限创意

引言 什么是Amazon Bedrock？ Amazon Bedrock是亚马逊云服务（AWS）推出的一项旗舰服务，旨在推动生成式人工智能（AI）在各行业的广泛应用。它的核心功能是提供由顶尖AI公司（如AI21 Labs、Anthropic、Cohere、Meta、Mistral AI、Stability AI以及亚马逊自身）开发的多种基础模型（Foundation Models，简称FMs）。
阅读更多...

Copyright China编程 23002807@qq.com

沪ICP备2023033072号-2