官术网_书友最值得收藏!

Using an optimized version of a Gaussian blur

A Gaussian blur is basically a convolution with a Gaussian function. Well, one of the features of convolutions is their associative property. This means that it does not matter whether we first invert the image and then blur it, or first blur the image and then invert it.

If we start with a blurred image and pass its inverse to the dodge function, then within that function the image will be inverted again (the 255-mask part), essentially yielding the original image. If we get rid of these redundant operations, the optimized convert_to_pencil_sketch function will look like this:

def convert_to_pencil_sketch(rgb_image):
gray_image = cv2.cvtColor(rgb_image, cv2.COLOR_RGB2GRAY)
blurred_image = cv2.GaussianBlur(gray_image, (21, 21), 0, 0)
gray_sketch = cv2.divide(gray_image, blurred_image, scale=256)
return cv2.cvtColor(gray_sketch, cv2.COLOR_GRAY2RGB)

For kicks and giggles, we want to lightly blend our transformed image (img_sketch) with a background image (canvas) that makes it look as though we drew the image on a canvas. So, before returning, we would like to blend with canvas if it exists:

    if canvas is not None:
gray_sketch = cv2.multiply(gray_sketch, canvas, scale=1 / 256)

We name our final function pencil_sketch_on_canvas, and it looks like this (together with optimizations):

def pencil_sketch_on_canvas(rgb_image, canvas=None):
gray_image = cv2.cvtColor(rgb_image, cv2.COLOR_RGB2GRAY)
blurred_image = cv2.GaussianBlur(gray_image, (21, 21), 0, 0)
gray_sketch = cv2.divide(gray_image, blurred_image, scale=256)
if canvas is not None:
gray_sketch = cv2.multiply(gray_sketch, canvas, scale=1 / 256)
return cv2.cvtColor(gray_sketch, cv2.COLOR_GRAY2RGB)

This is just our convert_to_pencil_sketch function, with the optional canvas argument that can add an artistic touch to the pencil sketch.

And we're done! The final output looks like this:

Let's take a look at how to generate a warming and cooling filter in the next section, where you'll learn how to use lookup tables for image manipulation.

主站蜘蛛池模板: 水城县| 馆陶县| 北票市| 荔波县| 阳江市| 原阳县| 霞浦县| 乾安县| 澜沧| 屏山县| 全椒县| 高邑县| 沅陵县| 九龙城区| 登封市| 太谷县| 秭归县| 灵丘县| 聊城市| 民权县| 大竹县| 榆中县| 柏乡县| 拜泉县| 鹤壁市| 山西省| 福安市| 贺州市| 汕尾市| 景谷| 大英县| 辰溪县| 福州市| 长沙县| 敖汉旗| 肇源县| 开封县| 杭锦后旗| 漠河县| 新竹县| 汕头市|