Abstract: A texture perception multimode coding for frame memory lossy compression is proposed to improve frame memory compression performance. First, the optimal directional reference pixel is calculated by using the texture perception and prediction residual is obtained by using the directional prediction. Then, rate-distortion is improved to obtain quantized parameter based on the continuity of motion direction and the correlation between quantization steps of same-position pixels between frames. Finally, according to the prediction residual characteristics of different texture regions, among the three encoding modes of run length coding, adaptive k Columbus coding and direct coding, adaptive selection of the optimal encoding mode is carried out. The simulation results show that, compared with the frame memory compression algorithm based on content-aware adaptive quantization, the average compression rate of this algorithm is improved by 14.8% when PSNR and encoding time are almost unchanged. The performance of the algorithm in this paper is strongly related to the complexity of the image, that is, the simpler the image texture is, the shorter the encoding time of the algorithm is, and the higher the compression rate is.