import cv2
from cv2.typing import MatLike
import numpy as np

POOL_SIZES = 10
POOL_KERNAL = np.ones((POOL_SIZES,POOL_SIZES),np.float32)/(POOL_SIZES*POOL_SIZES)

WINDOW_WIDTH = 21

def process_frame(frame: MatLike, thresh: float) -> tuple[MatLike, MatLike, list[tuple[int, int]], list[tuple[int, int]], list[tuple[int, int]]]:
    frame_gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
    frame_hls = cv2.cvtColor(frame, cv2.COLOR_BGR2HLS)
    H_filter_mask =  cv2.inRange(frame_hls[:,:,0], 100, 140)
    L_filter_mask =  cv2.inRange(frame_hls[:,:,1],  20, 210)
    S_filter_mask =  cv2.inRange(frame_hls[:,:,2], 200, 255)
    mask = cv2.bitwise_and(S_filter_mask, H_filter_mask)
    mask = cv2.bitwise_and(mask, L_filter_mask)
    frame_grey_filtered = cv2.bitwise_and(frame_gray, mask)
    frame_convolution_mask = cv2.inRange(cv2.filter2D(frame_grey_filtered, -1, POOL_KERNAL), 1, 255)

    WIDTH = frame_convolution_mask.shape[1]
    HEIGHT = frame_convolution_mask.shape[0]

    pos_windows = []
    neg_windows = []
    neutral_windows = []

    for x in range(0, WIDTH - WINDOW_WIDTH, 5):
        for y in range(0, HEIGHT - WINDOW_WIDTH, 5):
            score = np.sum(frame_convolution_mask[y:y+WINDOW_WIDTH, x:x+WINDOW_WIDTH])
            if score > 255*(WINDOW_WIDTH*WINDOW_WIDTH*thresh):
                pos_windows += [(x, y)]
            
            elif score < 1:
                neg_windows += [(x, y)]

            else:
                neutral_windows += [(x, y)]


    return frame_gray, frame_convolution_mask, pos_windows, neg_windows, neutral_windows