OpenCV: lossless capture?

Question

I wrote a simple python program that uses OpenCV and scantailor-cli to scan documents with my document camera and store them as high-contrast, low filesize PDFs.

Approach:

capture the document in full color with OpenCV at 1920x1080 in JPG format
convert to high-contrast black-and-white PDF with scantailor-cli

This generally works, however I have discovered an issue with my usage of OpenCV that is limiting the quality of my scans. I'm hoping someone here can give me the nudge I need to improve this.

When I use the UI-based capture software that came with my IPEVO document camera (Visualizer), I get a nice, clearn capture that makes a great conversion into high-contrast. When I use OpenCV, I get unwanted compression artifacts and conversion to high-contrast yields unsatisfactory results.

Full program listing is here: https://github.com/jonashw/Document-Scanner/blob/master/scan.py

Relevant excerpt below

def do_camera(device_id):
    cam = cv2.VideoCapture(device_id)#,cv2.CAP_DSHOW)
    w,h = 1920,1080
    cam.set(cv2.CAP_PROP_FRAME_WIDTH, w)
    cam.set(cv2.CAP_PROP_FRAME_HEIGHT, h)
    cam.set(cv2.CAP_PROP_FORMAT, 0)
    w = cam.get(cv2.CAP_PROP_FRAME_WIDTH)
    h = cam.get(cv2.CAP_PROP_FRAME_HEIGHT)
    grab_mode = cam.get(cv2.CAP_PROP_MODE)
    format = cam.get(cv2.CAP_PROP_FORMAT)
    print("capturing with w={}, h={}, grab_mode={}, format={}".format(w,h,grab_mode, format))
    
    #ok = False
    #i = -100
    #while (not ok) and i < 10:
        #if i != 0:
            #print("setting grab_mode {}".format(grab_mode + i))
            #ok = cam.set(cv2.CAP_PROP_MODE, grab_mode + i)
        #i += 1
    #if ok:
        #gm = cam.get(cv2.CAP_PROP_MODE)
        #printf("Grab mode = {}", format(gm))

    cv2.namedWindow("test", cv2.WINDOW_NORMAL)

    img_counter = 0
    img_paths = []

    while True:
        ret, frame = cam.read()
        if not ret:
            print("failed to grab frame")
            break
        cv2.imshow("test", frame)
        cv2.resizeWindow("test",1280,720)
        cv2.moveWindow("test",1920,0)

        k = cv2.waitKey(1)
        if k%256 == 27:
            # ESC pressed
            print("Escape hit, closing...")
            break
        elif k%256 == 32:
            # SPACE pressed
            img_name = "capture_{}.png".format(img_counter)
            img_path = os.path.join(capture_path, img_name)
            img_paths.append(img_path)
            os.makedirs(capture_path, exist_ok=True)
            cv2.imwrite(img_path, frame, [cv2.IMWRITE_PNG_COMPRESSION, 0])
            print("{} written!".format(img_name))
            img_counter += 1

    cam.release()

    cv2.destroyAllWindows()
    return img_paths

I suspect the exact line I need to alter is this one.

            cv2.imwrite(img_path, frame, [cv2.IMWRITE_PNG_COMPRESSION, 0])

As you can see, I have opted for PNG format, which should not result in any JPEG-like compression artifacts. Despite this selection, I get artifacts.

I have also tried embracing JPEG and setting the quality to max. This does not improve matters:

        cv2.imwrite(img_path, frame, [cv2.IMWRITE_JPEG_QUALITY, 100])

What can I do? Please and thank you

OpenCV: lossless capture?

0 Answers