Methods, systems and apparatuses for dual-camera based zooming

An apparatus for capturing and zooming an image by a desired zoom factor, comprising

• a dual camera unit captures a first image and a second image, wherein the first image is an image without zoom and the second image is an image with a predefined optical zoom;
• and
• a processing unit for
• determining a common area in the first image based on the second image;
• upscaling the determined common area in the first image by a predefined scaling factor for determining an upscaled version of the common area;
• dividing the upscaled version of the common area into a first set of overlapping patch vectors and calculating a magnitude of the first set of overlapping patch vectors by extracting the features of the first set of overlapping patch vectors and then storing the magnitude in a Low Resolution feature vector matrix;
• dividing the second image into a second set of overlapping patches for determining a second set of patch vectors and the second set of patch vectors is stored as a high resolution patch vector matrix;
• if the zoom factor is lesser than the predefined optical zoom, then
• upscaling the first image by the predefined optical zoom;
• dividing the upscaled first image into a third set of overlapping patch vectors and calculating a magnitude of the third set of overlapping patch vectors by extracting the features of the third set of overlapping patch vectors and then storing the magnitude in a first feature vectors matrix;
• determining one or more feature vectors of the upscaled first image that are similar to corresponding features stored in the Low Resolution feature vector matrix, wherein one or more features the upscaled first image are determined based on comparing the first feature vectors matrix and the Low Resolution feature vector matrix;
• determining a second set of patch vectors in the high resolution patch vector matrix corresponding to the one or more feature vectors of the upscaled first image;
• reconstructing a first set of estimated high resolution patch vectors based on the Gaussian weighting of the patch vectors in the second set of patch vectors;
• extracting a first high frequency content by subtracting the first image from an upscaled and downscaled version of the first image, wherein the upscaled and downscaled version of the first image is obtained by firstly upscaling the first image by a scaling factor and then downscaling the upscaled first image by the scaling factor;
• upscaling the first high frequency content by the scaling factor and adding a reconstructed zoomed version of the first image;
• if the zoom factor is greater than the predefined optical zoom, then
• upscaling the second image by the scaling factor;
• dividing the upscaled second image into a fourth set of overlapping patch vectors and calculating a magnitude of the fourth set of overlapping patch vectors by extracting the features of the fourth set of overlapping patch vectors and then storing the magnitude in a second feature vectors matrix;
• determining one or more feature vectors of the upscaled second image that are similar to corresponding features stored in the Low Resolution feature vector matrix, wherein one or more features the upscaled second image are determined based on comparing the second feature vectors matrix and the Low Resolution feature vector matrix;
• determining a third set of patch vectors in the high resolution patch vector matrix corresponding to the one or more feature vectors of the upscaled second image;
• reconstructing a second set of estimated high resolution patch vectors based on the Gaussian weight of the patch vectors in the third set of patch vectors;
• extracting a second high frequency content by subtracting the second image from an upscaled and downscaled version of the second image, wherein the upscaled and downscaled version of the second image is obtained by firstly upscaling the second image by a scaling factor and then downscaling the upscaled second image by the scaling factor;
• upscaling the second high frequency content by the scaling factor and and adding a reconstructed zoomed version of the second image.