Three-dimensional image coding method based on three-dimensional self-organized mapping

Abstract

The invention discloses a three-dimensional image coding method based on a three-dimensional self-organized mapping. The three-dimensional image coding method comprises the following steps: processing left and right images by a 3D SOM algorithm to obtain an optimal quantified code book; taking the left image as an input vector to carry out quantified coding; obtaining a quantification error by the left image and the quantified code vector; carrying out DCT conversion, quantification and Huffman coding; and reconstructing the right image by the reconstructed left image, a parallax error estimation error and a parallax error vector through a decoding end so as to realize effective coding of the three-dimensional images. According to the three-dimensional image coding method, a distance inequation is quoted to be used as judgment evidence and a frequency sensitive factor is also quoted; the three-dimensional self-organized mapping algorithm is improved so that the distortion measurement calculation complexity is reduced and the code book performance is improved; and the improved three-dimensional self-organized mapping algorithm is applied to three-dimensional image coding so that the coding frequency of the left image is effectively improved and the square effect of the reconstructed right image is reduced.

Description

technical field [0001] The invention relates to the field of image processing, in particular to a three-dimensional self-organizing map-based stereoscopic image coding method. Background technique [0002] Stereoscopic images increase the depth information of the scene and have a strong sense of presence and reality, so they are becoming more and more popular. The most widely used stereoscopic image is the binocular view solution, which uses the principle of binocular parallax to produce a stereoscopic effect. However, compared with two-dimensional plane images, the amount of stereoscopic image data has increased exponentially. Therefore, it is necessary to compress and encode stereoscopic images in order to effectively transmit and store them. [0003] In order to solve the above problems, the current main stereoscopic image coding methods include wavelet-based method, block-based method combined with MPEG (Moving Pictures Experts Group, Moving Pictures Experts Group) stand...

AI Technical Summary

Problems solved by technology

[0006] Method 3: The stereoscopic image coding method based on pattern recognition is a new image coding method. This method uses DE (Disparity Estimation, parallax estimation) compensation technology coding for the right image, and uses SOM (Self-organizing Map) for the left image. , self-organizing map) algorithm vector quantization coding to replace the traditional JPEG method, vector quantization and parallax estimation residuals are encoded using DCT (Discrete Cosine Transform, discrete cosine transform) + Huffman, this method can better It overcomes the shortcomings of insufficient adaptability of transform coding and has high compression efficiency, but it has high computational complexity and also has certain block effects

Images