Depth map estimation with 3DFFT for two-dimensional to three-dimensional stereoscopic conversion based on image registration

Due to the growing need for three-dimensional contents, the conversion of Two Dimensional (2D) images into Three Dimensional (3D) images has been a focus in 3D image processing. One of the most important enabling technologies for medical imaging and image-guided therapies is 3D image registration. Mass marketing is now hindered by 2D contents that need labour-intensive human editing of depth information, necessitating the creation of an effective 2D-to-3D conversion system. However, 3D image conversion with registration is quite challenging by using various existing approaches due to slow computation and small capture range. In order to overcome these challenges, depth map estimation with 3-Dimensional fast Fourier Transform (3DFFT) for stereoscopic conversion is developed. Initially, the left and right view of the x-ray images are collected and pre-processed using Adaptive Bilateral Filter (ABF) and Brightness Preserving Histogram Equalization (BBHE). ABF is used to reduce the image's noise, while BBHE is used to boost the picture's brightness. Then, features from the pre-processed image are identified using Speeded-Up Robust Feature (SURF). The left and right views of x-ray pictures features are matched and registered using 3DFFT. Finally, depth map estimation is designed in order to convert the registered image into a stereoscopic image. The automatic depth map estimation consists of four steps such as haze image simulation, first depth map extraction, refined map estimate and final depth map estimation. According to the experimental study, the proposed registration approach achieves 31.18dB of PSNR, 36dB of SNR, 0.05 of MSE and 0.40 of structural content. Thus the designed 3D registered model is the better option for real time analysis of bone structures.