No icon

Compression-dependent Transform Domain Downward Conversion for Block-based Image Coding

Compression-dependent Transform Domain Downward Conversion for Block-based Image Coding

Abstract:

Transform domain downward conversion (TDDC) for image coding is usually implemented by discarding some high-frequency components from each transformed block. As a result, a block of fewer coefficients is formed and a lower compression cost is achieved due to the coding of only a few low-frequency coefficients. In this paper, we focus on the design of a new TDDC-based coding method by using our proposed interpolation-compression directed filtering (ICDF) and error-compensated scalar quantization (ECSQ), leading to the compression-dependent TDDC (CDTDDC) based coding. More specifically, ICDF is first used to convert each 16 _ 16 macroblock into an 8_8 coefficient block. Then, this coefficient block is compressed with ECSQ, resulting in a smaller compression distortion for those pixels that locate at some specific positions of a macro-block. We select these positions according to the 4:1 uniform sub-sampling lattice and use the pixels locating at them to reconstruct the whole macro-block through an interpolation. The proposed CDTDDC-based coding can be applied to compress both grayscale and color images. More importantly, when it is used in the color image compression, it offers not only a new solution to reduce the data-size of chrominance components but also a higher compression efficiency. Experimental results demonstrate that applying our proposed CDTDDC-based coding to compress still images can achieve a significant quality gain over the existing compression methods.

Existing System:

In image coding, removing the spatial redundancy from the raw image data is always one of the top-priority considerations to reach a high compression. One straightforward way to achieve this goal is to perform the spatial sub-sampling on the source image to reduce its resolution, which accordingly produces a low-resolution (LR) image and forms the pixeldomain downward conversion (PDDC). Then, all compression operations are performed on the LR image to conduct the encoding. After the decoding is completed, an image interpolation is normally performed on the decoded LR image to reconstruct a full-resolution image. The decimation of the source image makes a lower bit-cost, which may offer a potential improvement on the compression efficiency. The coding efficiency of such a PDDC-based compression for grayscale images has been verified, where the LR image is generated by down-sampling the source image according to a uniform  sub-sampling lattice and the LR image is compressed by JPEG. This method has been effectively improved through introducing more sub-sampling modes to implement the spatial down-sampling for the compression.

 

Proposed System:

The image downward conversion may also be implemented in the transform domain by discarding some high-frequency components from the transformed block, which produces a small-sized coefficient block and yields the  transform domain downward conversion (TDDC). However, when the reconstruction is carried out, a serious quality degradation often happens due to the lack of necessary high-frequency information. This also limits the application of TDDC in practical image coding.

The design of IDID aims at a high-efficient interpolation by preserving enough prior information of the source image. However, after such a down-sampling, the pixels within a down-sized block are not so closely correlated as before. Therefore, performing the transform coding on the downsized block cannot guarantee a high compression efficiency, often resulting in a higher bit-cost. To solve this problem, we propose a new down-sampling method to reduce the image data-size in the transform domain rather than in the pixel domain. The most important advantage of our proposed down-sampling is that it not only transfers the high-frequency information from the source image to the down-sized image but also guarantees a lower bit-cost for the compression.

Comment As:

Comment (0)