Advanced one-dimensional triple wavelet analysis in row for infrared images from un-cooled infrared MEMS system

For the limitation of detecting materials, the images from the novel un-cooled infrared system based on visible light readout are blurry and have low contrast. The images also have more noise and larger holes. Especially after pseudo-color processing, the noise and holes will become much clearer. For the characteristics of images in the un-cooled IR system, the traditional image processing methods for IR images are not suitable for the image in our research. Therefore, an advanced one-dimensional triple wavelet analysis in row for infrared images is presented based on the characteristics of un-cooled infrared system. In this method, the triple wavelet decomposition is made in each row of image, and detail coefficients and approximation coefficients of each row can be obtained. The detail coefficients in the first time of wavelet decomposition express the whole details of image containing noise and the edge of object. So after low-pass filter, the noise in the image can be suppressed. By the wave reconstruction made between the approximation coefficients in triple wavelet decomposition and the detail coefficients after low-pass filter, each row in images without noise and holes can be gained. In wavelet reconstruction, a weight being proportional with the filtering window is multiplied with detail coefficients. The weight can make sure the gray value of whole picture and the contrast cannot be lower after low-pass filter. The images from un-cooled infrared system are processed in the computer with the software of MATLAB. The results support that compared with traditional methods the novel method can be more effective to eliminate the noise and fill holes, and better response to the temperature details of objects.