当前位置: 首页 > 专利查询>清华大学专利>正文

编码方法、解码方法及其装置制造方法及图纸

技术编号:17963901 阅读:28 留言:0更新日期:2018-05-16 07:15
一种图像编码方法、解码方法以及对应装置和帧内像素预测方法。所述图像编码方法包括:像素分块步骤,将一帧图像中的像素分块;像素值预测步骤,对于当前待编码块,对于块中的每个像素,分别进行像素值预测处理,得到第一预测值;残差计算步骤,对于当前待编码块,对于块中的每个像素,计算残差;离散余弦变换、量化和熵编码步骤,对于残差计算步骤得到的当前待编码块对应的残差块进行离散余弦变换、量化和熵编码,其中量化后的残差块用于在编码端的像素值重构步骤,熵编码后的残差块供发送到解码端;像素值重构步骤,对于先前已编码的像素的像素值进行重构,重构的像素值用于像素值预测步骤;残差计算步骤包括:基于像素值预测步骤预测得到的该像素的第一预测值和该像素的真实值,计算关于该像素的第一残差;确定该像素的第一残差是正常残差还是异常残差;如果该像素的第一残差为异常残差,则进行异常残差修正,包括记录该出现异常残差的像素的位置;对该像素进行重新预测,得到第二预测值和对应的第二残差,并记录重新预测信息,使得编码端的像素值重构步骤和解码端都能够基于该出现异常残差的像素的位置和重新预测信息,重构和解码该像素的像素值;其中由经过异常残差修正的、该当前待编码块的像素的残差组成该当前待编码块的正常残差块,供离散余弦变换、量化和熵编码步骤进行下一步的处理,将出现异常残差的像素的位置和重新预测信息单独进行编码,该编码过程是无损的,经编码后的像素的位置和重新预测信息用于像素值重构步骤和用于发送到解码端。可以在像素分块之前,对像素进行重新排列,由此避免误差扩散。在像素值预测处理中,可以先对预测算法进行预测,然后根据确定的预测算法进行像素值预测。通过将像素残差分为正常残差和异常残差,以及对于异常残差的像素进行重新预测,能够提高图像编解码的准确性。

Coding method, decoding method and its device

The invention relates to an image coding method, a decoding method, a corresponding device and a pixel prediction method in the frame. The image coding method includes pixel block step, pixel block in one frame image, pixel value prediction step, pixel value prediction processing for each pixel in block, and first prediction value for each pixel in block. The discrete cosine transform, quantization and entropy coding steps are used to make discrete cosine transform, quantization and entropy coding for the residual block corresponding to the current code block obtained by the residual calculation step, in which the quantized residual block is used to reconstruct the pixel value at the coding end, and the residual block after the entropy coding is sent for transmission. To the decode; the pixel value reconfiguration step is reconstructing the pixel value of the previously encoded pixel, the reconstructed pixel value is used for the pixel value prediction step, and the residual calculation step includes the first prediction value of the pixel and the true value of the pixel predicted based on the pixel value prediction step, and the first residue of the pixel is calculated. Difference; determines whether the first residual difference of the pixel is a normal residual or an abnormal residual; if the first residual difference of the pixel is an abnormal residual, the abnormal residual correction is made, including the location of the pixel that records the abnormal residual, and the pixel is repredicted, the second prediction value and the corresponding second residuals are obtained, and the record weight is recorded. The new prediction information enables the pixel value reconstruction step and the decode end of the coding end to reconstruct and decode the pixel values based on the location and reprediction of the pixels that have the abnormal residuals, in which the residuals of the pixels of the currently pending block, modified by the abnormal residuals, constitute the positive block. A constant residual block, which is processed for the next step of the discrete cosine transform, quantization and entropy coding steps, encoding the position of the pixels of the abnormal residual and the reprediction information separately, the encoding process is lossless, the encoded pixel position and the reprediction information are used in the pixel value reconstruction step and used to send to the code. Decode. Pixels can be rearranged before pixels are partitioned, so as to avoid error diffusion. In the pixel value prediction processing, the prediction algorithm can be predicted first, and then the pixel value is predicted according to the prediction algorithm. The accuracy of the image codec can be improved by dividing the pixel residuals into normal residuals and abnormal residuals, as well as the re prediction of the pixels of the abnormal residual.

【技术实现步骤摘要】
【国外来华专利技术】PCT国内申请,说明书已公开。

【技术保护点】
PCT国内申请,权利要求书已公开。

【技术特征摘要】
【国外来华专利技术】PCT国内申请,...

【专利技术属性】
技术研发人员:陶品冯立新王正
申请(专利权)人:清华大学
类型:发明
国别省市:北京,11

网友询问留言 已有0条评论
  • 还没有人留言评论。发表了对其他浏览者有用的留言会获得科技券。

1