Method for image cropping
View Patent ↗A method to crop image from an irregularly shaped region. A top-left corner of the best cropping rectangle are determined by sequentially scanning the rows of the irregularly shaped region to find out the largest bounding area. By rolling back the largest bonding area from the x or y direction, the bottom-right corner is determined.
1. A method for image cropping comprising the steps of:
providing a plurality of images, wherein the images overlap with each other to generate an irregularly shaped region composed of columns and rows of pixels, wherein the pixels of the irregularly shaped region are elements of a pixel conjunction of transformed pixel sets of a plurality the images;
automatically locating a cropping rectangle within the irregularly shaped region by automatically searching for a first corner and a second corner of the cropping rectangle, wherein the first corner and the second corner form the cropping rectangle, and the search method for the first corner and the second corner comprises the steps of:
calculating, within the irregularly shaped region for each of the rows, a first width between a left and right boundary pixel of the row, a height between the left and a boundary pixel of the same column, another height between the left and the other boundary pixel of the same column, and first products of the widths and the heights;
identifying one of the left boundary pixels in the irregularly shaped region as the first corner of the cropping rectangle, wherein the first product calculated for the row of the first corner is the largest, and a semi-finished rectangle region is formed by the first corner, and the length and width which yield the largest first product;
calculating, within an overlapping region of the semi-finished rectangle region and the irregularly shaped region, for each of the rows, a width between a right and left boundary pixel of the row, a height between the right and a boundary pixel of the same column, another height between the right and the other boundary pixel of the same column, and second products of the widths and the heights;
identifying one of the right boundary pixels in the overlapping region as the second corner of the cropping rectangle, wherein the second product calculated for the row of the second corner is the largest; and
cropping the cropping rectangle with the first corner and the second corner from the images.
2. The method as claimed in claim 1 , wherein the calculations of the first products are implemented for the rows in two sequences from top to bottom and from bottom to top.
3. The method as claimed in claim 1 , wherein the calculations of the second products are implemented for the rows in a sequence from bottom to top.
4. The method as claimed in claim 1 , wherein the calculations of the second products are implemented for the rows in a sequence from top to bottom.
5. The method as claimed in claim 1 further comprising creating a mask distinguishing the irregularly shaped region.
6. The method as claimed in claim 5 , wherein column indices of the left and right boundary pixels are stored in a one-dimension edge list array having a length the same as a height of the mask.