Acceleration method for realizing inverse wavelet transform in JPEG2000 decompression algorithm by using GPU

Abstract

The invention discloses an acceleration method for realizing inverse wavelet transform in an JPEG2000 decompression algorithm by using a GPU. The method comprises the following steps: firstly, performing optimized storage for inverse wavelet transform coefficients in a shard memory by using the advantage that the GPU shared memory is high in access speed; and secondly, adopting a parallel processing mode that one block processes a line (column) of pixels according to the characteristics that the line-line processing and column-column processing in the inverse wavelet transform process are independent of each other, adopting a parallel processing mode that one thread processes a plurality of pixels according to the characteristics that the processing of the same line (column) of pixels has dependence, but part of the processing procedure can still be parallel, and completing the inverse wavelet transform in a horizontal direction and a vertical direction. According to the inverse wavelet transform acceleration method disclosed by the invention, the advantages that the GPU has a large number of cores and is suitable for parallel computation are fully used, and the decompression speed of JEPG2000 can be improved effectively; and moreover, the algorithm structure is simple and reasonable, and is easy to implement on the GPU.

Description

technical field [0001] The invention belongs to the field of multimedia encoding and decoding, and in particular relates to an acceleration method for implementing inverse wavelet transform in JPEG2000 decompression algorithm by using GPU. Background technique [0002] With the rapid development of multimedia technology and network communication technology, people urgently need a static image compression standard with better compression effect, more image formats and more new features. In December 2000, the Joint Photographic Experts Group (JPEG) launched a new generation of still image compression standard—JPEG2000. JPEG2000 adopts wavelet transform, optimized truncation embedded block coding and efficient code stream organization form, which is suitable for various types of static images. The coding system flow of JPGE2000 is as follows: figure 1 shown. In the encoding process, the image is first preprocessed (including image block, level shift and component transformat...

AI Technical Summary

Problems solved by technology

[0005] The excellent compression performance of JPEG2000 is achieved through a highly complex algorithm, and the complexity of its decompression algorithm is relatively high, which restricts the application of JPEG2000 to a certain extent, especially in systems with high real-time requirements

Images