Method and device for minimizing a quantization error

#This application is based on and hereby claims priority to International German Application No. 10 2005 016 858.2 filed on Apr. 12, 2005, the contents of which are hereby incorporated by reference.

#When compressing information a distinction is made between lossless and lossy compression techniques. With lossless compression techniques, such as Lempel-Ziv coding or Huffman coding, the original, non-compressed information can be reconstructed from compressed information without errors. Methods of this kind only achieve a high compression rate if the information for compressing comprises specific structures. By contrast high compression rates can be achieved with lossy compression techniques, such as the JPEG method (JPEG-Joint Picture Expert Group), the MPEG2 method (MPEG-Motion Picture Expert Group) or the MC method (MC-Adaptive Audio Coding). However it has to be accepted in this connection that the original, non-compressed information cannot be recovered from the compressed information error free. The errors in the decompressed information often increases as the compression rate increases. Thus with the image compression method JPEG for example, block artifacts can be seen within the decompressed image.

#FIG. 1 exemplarily shows a simplified encoding unit of a JPEG encoder JC. In this connection an image IM for compressing is divided into image blocks BB with, for example, 8×8 pixels. Further processing within the JPEG encoder JC is based on these image blocks. Each image block is firstly transformed by the discrete cosine transformation FDCT into the frequency. The coefficients X are produced in the process. Each coefficient X is subjected to a quantization FQ, the quantization, i.e. the quantization factor Q, being controlled by the first table TS1. The quantized coefficient Z is generated hereby. Entropy coding then takes place, for example Huffman coding, by an entropy encoder EC which is controlled with the aid of a second table TS2. Coded image data is written into a file JDS at the outcome of entropy coding.

#These and other objects and advantages of the present invention will become more apparent and more readily appreciated from the following description of the preferred embodiments, taken in conjunction with the accompanying drawings of which:

#FIG. 1 shows a schematic construction of a JPEG encoder of the related art,

#FIG. 2 shows a schematic construction of a JPEG decoder of the related art,

#Reference will now be made in detail to the preferred embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein like reference numerals refer to like elements throughout.

#FIGS. 1 and 2 have been acknowledged in detail in the introduction, so they will not be discussed further at this point.

#The method will be described in more detail with the aid of FIGS. 3 and 5. FIG. 3 is based on FIG. 1, having been supplemented by a correction value unit KWE. In FIG. 3 compressed data is generated from an image IM, which data is stored in a first file JDS1. In the quantization with the quantization factor Q a quantized coefficient is generated for each coefficient X. In the process the quantization factor Q, the coefficients X and the quantized coefficients Z are fed to the correction value unit KWE as the set of coefficients. The correction value unit KWE stores for example a first quantization error QF1 which is produced from QF1=(X−Z*Q)2. Furthermore, the quantization factor Q used in the process is stored for each first quantization error QF1. The following table shows this by way of example using five sets of coefficients, these sets of coefficients being numbered with a number from to 1 to 5: