Based on bit plane and adaptive binary arithmetic coding, a wavelet coefficients coding scheme which could match the line-based wavelet transform efficiently was proposed. The MCU of the image sensor node read image data from FIFO line by line, executed multi-level wavelet transform, determined the quantization value according to the probability distribution of the wavelet coefficients in different level, performed adaptive binary arithmetic coding based on four binary probabilistic model and realized image compression based on wavelet transform with a low cost of SRAM. The image compression method was applied to handle a 320 pixels×240 pixels gray image of piglets. Experimental results show that the SRAM cost, time cost and the PSNR was 5749KB, 16312s and 39.72dB respectively when the quantization value was set to three. This study established the foundation for agricultural image transmission over the low bandwidth WMSN efficiently.