Permalink
Cannot retrieve contributors at this time
Fetching contributors…
Cannot retrieve contributors at this time
| import cv2, numpy as np, draw_utils, image_utils | |
| from time import time | |
| __winname = "Eye Centre (gradients)" | |
| __debug_imgs = {} | |
| __inv_intensity_weight_divisor = 100 | |
| __fast_width = 25.0 | |
| # Algorithm from "ACCURATE EYE CENTRE LOCALISATION BY MEANS OF GRADIENTS - Fabian Timm and Erhardt Barth" | |
| # Based on C++ code from https://github.com/trishume/eyeLike | |
| def test_possible_centers_formula((grad_x0, grad_y0), darkness_weight, grad_x_val, grad_y_val, indicies_grid, shape): | |
| dx = np.ones(shape) * grad_x0 - indicies_grid[1] | |
| dy = np.ones(shape) * grad_y0 - indicies_grid[0] | |
| magnitudes = cv2.magnitude(dx + 0.0001, dy) # 0.0001 is a hack to offset against division by 0 | |
| dx = dx / magnitudes | |
| dy = dy / magnitudes | |
| diffs = (dx * grad_x_val + dy * grad_y_val) * darkness_weight | |
| diffs[diffs < 0] = 0 | |
| return diffs | |
| def get_center_map(eye_img_grey): | |
| grad_x_img = cv2.Sobel(eye_img_grey, ddepth=cv2.CV_32F, dx=1, dy=0, ksize=5) | |
| grad_y_img = cv2.Sobel(eye_img_grey, ddepth=cv2.CV_32F, dx=0, dy=1, ksize=5) | |
| magnitudes = np.sqrt(grad_x_img ** 2 + grad_y_img ** 2).astype(int) | |
| mag_thresh = int(np.std(magnitudes) / 2 + np.mean(magnitudes)) | |
| grad_x_img = np.divide(grad_x_img, magnitudes + 1) | |
| grad_y_img = np.divide(grad_y_img, magnitudes + 1) | |
| grad_x_img[magnitudes < mag_thresh] = 0 | |
| grad_y_img[magnitudes < mag_thresh] = 0 | |
| # inverted image for weighting | |
| eye_img_inv = 255 - eye_img_grey | |
| darkness_weights = eye_img_inv / __inv_intensity_weight_divisor | |
| accumulator = np.zeros(eye_img_grey.shape[:2], dtype=np.float32) | |
| indicies_grid = np.indices(accumulator.shape[:2]) | |
| indicies_shape = indicies_grid.shape[1:3] | |
| for y in range(eye_img_grey.shape[0]): | |
| for x in range (eye_img_grey.shape[1]): | |
| if grad_x_img[y][x] == 0 and grad_y_img[y][x] == 0: continue | |
| accumulator += test_possible_centers_formula((x, y), | |
| darkness_weights[y][x], | |
| grad_x_img[y][x], grad_y_img[y][x], | |
| indicies_grid, indicies_shape) | |
| num_gradients = eye_img_grey.shape[0] * eye_img_grey.shape[1] | |
| return accumulator / num_gradients | |
| def find_pupil(eye_img_bgr, debug_index=False): | |
| """ Estimates the centre of the pupil using image gradients | |
| """ | |
| eye_img_r = cv2.split(eye_img_bgr)[2] # Extract red channel only | |
| # Scale to small image for faster computation | |
| scale = __fast_width / eye_img_bgr.shape[0] | |
| small_size = (int((__fast_width / eye_img_bgr.shape[0]) * eye_img_bgr.shape[1]), int(__fast_width)) | |
| eye_img_small = cv2.resize(eye_img_r, small_size) | |
| center_map = get_center_map(eye_img_small) | |
| max_val_index = np.argmax(center_map) | |
| pupil_y0, pupil_x0 = max_val_index // center_map.shape[1], max_val_index % center_map.shape[1] | |
| # Scale back to original coordinates | |
| pupil_y0, pupil_x0 = int((pupil_y0 + 0.5) / scale), int((pupil_x0 + 0.5) / scale) | |
| if debug_index: | |
| eye_img_r_debug = cv2.cvtColor(eye_img_r, cv2.COLOR_GRAY2BGR) | |
| debug_img = eye_img_bgr.copy() | |
| cmap_norm = cv2.normalize(center_map, alpha=0, beta=255, norm_type=cv2.NORM_MINMAX) | |
| center_map_big = cv2.resize(cmap_norm, (eye_img_bgr.shape[1], eye_img_bgr.shape[0])) | |
| center_map_big = cv2.cvtColor(center_map_big.astype(np.uint8), cv2.COLOR_GRAY2BGR) | |
| draw_utils.draw_cross(debug_img, (pupil_x0, pupil_y0), (0, 255, 255), 6) | |
| draw_utils.draw_cross(center_map_big, (pupil_x0, pupil_y0), (255, 0, 0), 6) | |
| stacked_imgs = image_utils.stack_imgs_horizontal([debug_img, eye_img_r_debug, center_map_big]) | |
| __debug_imgs[debug_index] = stacked_imgs | |
| if debug_index == 2: | |
| full_debug_img = image_utils.stack_imgs_vertical([__debug_imgs[1], __debug_imgs[2]]); | |
| cv2.imshow(__winname, full_debug_img) | |
| elif debug_index > 2: | |
| cv2.imshow(__winname, stacked_imgs); | |
| return pupil_x0, pupil_y0 | |
| #---------------------------------------- | |
| # EXAMPLE USAGE | |
| #---------------------------------------- | |
| if __name__ == '__main__': | |
| eye_img = cv2.imread('eye_images/phil1_l.png', 3) | |
| tic = time() | |
| for a in range(100): | |
| (pupil_x0, pupil_y0) = find_pupil(eye_img, debug_index=3) | |
| print 'Time taken to find pupil: %0.3f s' % (time() - tic) | |
| cv2.waitKey() | |