routes/sudoku.py

from flask import Flask, request, jsonify
import base64
import io
from PIL import Image
import cv2
import numpy as np
import easyocr
import logging

from routes import app

# Configure logging
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)

# Initialize EasyOCR Reader
reader = easyocr.Reader(['en'], gpu=False)

def run_length_decode(encoded_bytes):
    if len(encoded_bytes) % 2 != 0:
        logger.error("Run-length encoded data length is not even.")
    
    decoded = bytearray()
    for i in range(0, len(encoded_bytes), 2):
        count = encoded_bytes[i]
        byte = encoded_bytes[i + 1]
        decoded.extend([byte] * count)
    
    return bytes(decoded)

def remove_green_lines(cv_image):
    hsv = cv2.cvtColor(cv_image, cv2.COLOR_BGR2HSV)
    lower_green = np.array([40, 40, 40])
    upper_green = np.array([80, 255, 255])
    mask = cv2.inRange(hsv, lower_green, upper_green)
    mask_inv = cv2.bitwise_not(mask)
    img_no_green = cv2.bitwise_and(cv_image, cv_image, mask=mask_inv)
    return img_no_green

def preprocess_image(cv_image):
    gray = cv2.cvtColor(cv_image, cv2.COLOR_BGR2GRAY)
    blurred = cv2.GaussianBlur(gray, (7, 7), 0)
    thresh = cv2.adaptiveThreshold(
        blurred, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C,
        cv2.THRESH_BINARY, blockSize=15, C=3)
    thresh = cv2.bitwise_not(thresh)
    kernel = np.ones((5,5), np.uint8)
    thresh = cv2.morphologyEx(thresh, cv2.MORPH_CLOSE, kernel)
    return thresh

def extract_sudoku_board(image, emptyCells, image_size):
    cv_image = cv2.cvtColor(np.array(image), cv2.COLOR_RGB2BGR)
    cv_image = remove_green_lines(cv_image)
    thresh = preprocess_image(cv_image)
    contours, _ = cv2.findContours(thresh, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
    
    largest_contour = max(contours, key=cv2.contourArea)
    peri = cv2.arcLength(largest_contour, True)
    epsilon = 0.01 * peri
    approx = cv2.approxPolyDP(largest_contour, epsilon, True)

    pts = approx.reshape(4, 2)
    rect = np.zeros((4, 2), dtype="float32")
    s = pts.sum(axis=1)
    rect[0] = pts[np.argmin(s)]
    rect[2] = pts[np.argmax(s)]
    diff = np.diff(pts, axis=1)
    rect[1] = pts[np.argmin(diff)]
    rect[3] = pts[np.argmax(diff)]

    (tl, tr, br, bl) = rect
    widthA = np.linalg.norm(br - bl)
    widthB = np.linalg.norm(tr - tl)
    maxWidth = max(int(widthA), int(widthB))
    heightA = np.linalg.norm(tr - br)
    heightB = np.linalg.norm(tl - bl)
    maxHeight = max(int(heightA), int(heightB))

    dst = np.array([
        [0, 0],
        [maxWidth - 1, 0],
        [maxWidth - 1, maxHeight - 1],
        [0, maxHeight - 1]
    ], dtype="float32")

    M = cv2.getPerspectiveTransform(rect, dst)
    warp = cv2.warpPerspective(cv_image, M, (maxWidth, maxHeight))

    # Determine Sudoku size based on image dimensions and emptyCells
    size = 4  # Default to 4x4
    if image_size > 40000:
        if image_size < 100000:
            size = 9
        else:
            size = 16
    
    # Secondary check based on emptyCells
    empty_cell_check = any(cell.get('x', 0) > 9 or cell.get('y', 0) > 9 for cell in emptyCells)
    if empty_cell_check:
        size = 16

    # Calculate cell size
    cell_size = maxWidth // size
    
    # Resize the warped grid to the expected size
    expected_size = size * cell_size
    warp = cv2.resize(warp, (expected_size, expected_size))

    # Convert warped image to grayscale and threshold
    warp_gray = cv2.cvtColor(warp, cv2.COLOR_BGR2GRAY)
    warp_blurred = cv2.GaussianBlur(warp_gray, (5, 5), 0)
    warp_thresh = cv2.adaptiveThreshold(
        warp_blurred, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C,
        cv2.THRESH_BINARY, 11, 2)
    warp_thresh = cv2.bitwise_not(warp_thresh)

    board = []
    for y in range(size):
        row = []
        for x in range(size):
            start_x = x * cell_size
            start_y = y * cell_size
            end_x = (x + 1) * cell_size
            end_y = (y + 1) * cell_size
            
            cell = warp_thresh[start_y:end_y, start_x:end_x]
            digit = recognize_digit(cell, size)
            row.append(digit)
        board.append(row)
    
    return board, size

def count_connected_black_pixels(image, min_pixels=14):
    # Label the connected components in the binary image
    image = cv2.resize(image, (100, 100))

    num_labels, labels_im = cv2.connectedComponents(image, connectivity=8)
    
    # Save Image for inspection in img folder with randome name
    # cv2.imwrite("img/img" + str(np.random.randint(0, 1000)) + ".jpg", image)

    # Subtract 1 to ignore the background label (label=0 is the background)
    valid_components = 0
    
    # Iterate over all the labels and count those with at least min_pixels pixels
    for label in range(0, num_labels):
        if np.sum(labels_im == label) >= min_pixels:
            valid_components += 1
    
    valid_components = valid_components - 2

    if valid_components < 0:
        valid_components = 0
    
    return valid_components

# Test the function on a preprocessed image
def recognize_digit(cell_image, slots = 4):
    """
    Recognizes the digit in a Sudoku cell. If no digit is detected, it counts black pixel groups (dots).
    """
    # Preprocess cell image for digit detection
    if slots == 4:
        cell_resized = cv2.resize(cell_image, (100, 100))
    elif slots == 9:
        cell_resized = cv2.resize(cell_image, (300, 300))
    else:
        cell_resized = cv2.resize(cell_image, (500, 500))
    cell_padded = cv2.copyMakeBorder(
        cell_resized, 10, 10, 10, 10, cv2.BORDER_CONSTANT, value=255)
    cell_rgb = cv2.cvtColor(cell_padded, cv2.COLOR_GRAY2RGB)    

    # Use EasyOCR to detect text
    result = reader.readtext(cell_rgb, detail=0, allowlist='0123456789')
    
    if result:
        digit = max(result, key=len)
        if digit.isdigit():
            return int(digit)
    
    # If no digit is detected, count connected black pixel components (dots)
    logger.info("No digit detected. Counting black pixel groups.")
    # Save Image for inspection
    dots_count = count_connected_black_pixels(cell_padded)
    
    logger.info(f"Detected {dots_count} connected black pixel groups (dots).")
    
    return dots_count if dots_count > 0 else 0  # Default to 0 if nothing is recognized

def is_valid_general(board, row, col, num, size):
    # Check row and column
    for i in range(size):
        if board[row][i] == num or board[i][col] == num:
            return False
    # Check subgrid
    subgrid_size = int(np.sqrt(size))
    start_row, start_col = subgrid_size * (row // subgrid_size), subgrid_size * (col // subgrid_size)
    for y in range(start_row, start_row + subgrid_size):
        for x in range(start_col, start_col + subgrid_size):
            if board[y][x] == num:
                return False
    return True

def solve_sudoku_general(board, size):
    for row in range(size):
        for col in range(size):
            if board[row][col] == 0:
                for num in range(1, size + 1):
                    if is_valid_general(board, row, col, num, size):
                        board[row][col] = num
                        if solve_sudoku_general(board, size):
                            return True
                        board[row][col] = 0
                return False
    return True

def calculate_sum(board, empty_cells):
    total = 0
    for cell in empty_cells:
        x = cell.get('x')
        y = cell.get('y')
        
        # Validate the coordinates

        total += board[y][x]
    
    return total

@app.route('/sudoku', methods=['POST'])
def sudoku_endpoint():
    payload = request.json

    logger.info("Received payload: %s", payload)
    
    # Extract fields
    sudoku_id = payload.get('id')
    encoded_str = payload.get('encoded')
    img_length = payload.get('imgLength')
    empty_cells = payload.get('emptyCells')
    
    # Base64 Decode First
    run_length_encoded_bytes = base64.b64decode(encoded_str)
    logger.info("Base64 decoding successful.")
    
    # Run-Length Decode Second
    image_bytes = run_length_decode(run_length_encoded_bytes)
    logger.info("Run-length decoding successful.")
    
    # Open image using PIL
    image = Image.open(io.BytesIO(image_bytes))
    # Save Image for inspection
    # image.save("image_main.jpg")
    logger.info("Image opened successfully.")
    
    # Extract Sudoku board

    board, size = extract_sudoku_board(image, empty_cells, img_length)
    logger.info(f"Sudoku board extracted successfully. Size: {size}x{size}")
    
    # Solve Sudoku

    logger.info("Sudoku board before solving:")
    for row in board:
        logger.info(row)
    solved = solve_sudoku_general(board, size)
    logger.info("Sudoku solved successfully.")
    
    total_sum = calculate_sum(board, empty_cells)
    logger.info(f"Sum of specified cells: {total_sum}")
    
    # Prepare response
    response = {
        "answer": board,
        "sum": total_sum
    }
    
    logger.info(f"Response prepared successfully for Sudoku ID: {sudoku_id}")
    
    return jsonify(response)