The invention belongs to the field of image compression, and provides an image compression sensing method based on a JND model. The method is used for improving the efficiency of an image compressionsensing algorithm. The method comprises the steps: building a sensing error threshold value through the JND model, and dividing an original image block into a human eye sensitive image block and a human eye insensitive image block; distributing more samples for the human eye sensitive image block in a sample distribution process of compression sensing sampling, distributing fewer samples for the human eye insensitive image block, adaptively distributing samples according to the sensitivity degrees of human eyes to the internal features of an image under the condition that the overall samplingrate is not changed, thereby achieving the reasonable distribution of samples, improving the sampling efficiency, and improving the reconstruction quality of the image.