The steganography that we suggested uses the data that ties strongly to the characteristics of the original image as watermark information without changing the original image. Although similarity with zero-watermarking is pointed out, the major difference is that our method does not need pre-processing for a feature extraction at decoding. Since the decoding can do at high speed, besides using as the original steganography, it is possible to apply it to similar image retrieving, similar image order sorting or template matching. For example, when applying it to medical images, a doctor is possible to pick some similar images from a medical image database and refers the treatments through the images with concealing the information of the patient. This time, we have investigated the identification accuracy of our method with respect to the template matching of similar image searching.