This paper adopts a line based wavelet transform method with the Le Gall 5/3 filter, in which the image lines are read to SRAM line by line from the SD card and write to the SD card after the multi-level transform operation. The wavelet transform method requires low SRAM memory and involves fixed point integer multiplication, addition, shift operation only. Appling the method to a 256 pixels ×256 pixels gray image of piglets. experimental results show that the SRAM cost of the method is 3.968KB and the time cost is 8.718s with a reasonable cost of the arithmetic operation. This study establishes the foundation for the wavelet transform image compression on the image sensor nodes, and makes it possible to transmit the images of agriculture over the low bandwidth WMSN efficiently.