Research On Shape-adaptive Transform And Its Application In Region Of Interest Coding

In recent years, region of interest (ROI) coding is a kind of up-and-coming technology of image compression. At low bit rate, region of interest are assigned to more bits, but the background is used high compression ratio to reduce the total amount of data, while the quality of whole image can meet the requirements of the observer. Generally, ROI is arbitrary shape, but the traditional DCT can only applies to rectangular block. Thus, the objective of this paper is to research the shape-adaptive transform algorithm of the ROI in image coding, contained shape-adaptive DCT (SA-DCT) and shape-adaptive APBT (SA-APBT)Orthogonal transform has been widely applied to image coding, video compression. A new member in the family of BTs, named all phase biorthogonal transform (APBT), has showed the good performance in image coding. This paper introduces all phase (discrete cosine / inverse discrete cosine) biorthogonal transforms. APBT can be used as the new transform instead of DCT to implement the well-known JPEG scheme, adopted uniform quantization. Compared with the traditional JPEG algorithm, better subjective effects and subjective quality are achieved at low bit rate.This paper introduces the shape-adaptive transform algorithm of ROI coding. Then, the experiments and analysis are as follows: (1) Set different sizes of circle shaped ROI of typical image, used SA-DCT to the ROI and background region in different compression ratio. Analyze the influence of area ratio of ROI in entire map. (2) Code the circle and arbitrary shaped ROI in SA-DCT, it shows that SA-DCT has a notable advantage compared to conventional DCT transform in the ROI. (3) Combined with SA-DCT and APBT algorithm innovatively, proposed SA-APBT, used it in the coding framework of JPEG, the coding quality can be improved slightly at low bit rate. In addition, another advantage of this algorithm is that the uniform quantization is used instead of complex quantization table. (4) Combined with SA-DCT and SA-APBT algorithm innovatively, proposed a new ROI coding scheme, used it to the ROI coding experiments of the entire image. This algorithm provides the ROI of image much higher coding quality than the JPEG, and can achieve a good visual quality.