Python - Image from numpy array losing true pixels

Question

I have a jpg picture of a face, I need to access the picture pixel by pixel (know what value is at each pixel), and use some sort of DFS to change background color.

image = Image.open("pic.jpg")
image = np.array(image)

First of all, why is the shape of the array (473, 354, 3)? It doesn't make sense to me.

When I do

plt.imshow(image.reshape(473, -1))
plt.show()

I get a picture that looks like the following, which consists of only red, blue and yellow colors (and a mixture of the three?)

This means that the values in the array are not what I can reliably use to make my edge detection decisions.

Why and what should I do?

I want the pixel values to reflect the true color of the original image, not like above.

The background in the actual picture is kinda white, and I want them and all other pixel values to stay that way, so I can implement my algorithm.

Charles McGuinness · Accepted Answer · 2016-05-17 08:06:04Z

1

The 3 is because each color (blue, green red) gets its own entry in the array.

For edge detection, you would might do best to collapse the image down to B&W. OpenCV has cv2.cvtColor(image, cv2.COLOR_BGR2GRAY) that will do the trick.

answered May 17, 2016 at 8:06

Charles McGuinness

212 bronze badges

Sign up to request clarification or add additional context in comments.

4 Comments

Jobs Over a year ago

So (473, 354, 3) means: the height is 473, the width is 354, and each value in the array is a 3 - tuple?

syntonym Over a year ago

@Jobs exactly. Also the colors that are shown are not the colors in the real picture but it's a colormap that shows specifici colors for high values. So if you input the picture with dimensions (h,w,(RGB)) probably only the red value is taken into account. It will show red where high values (of red) are and blue where low values (of red) are. You probably should convert to black&white and don't take the colors literally (or switch to a black and white colormap).

Jobs Over a year ago

Thank you so much! That makes sense. So now I've successfully accomplished the task - changing the background of my picture to white. The size of the picture dropped from 60 kb to 20 kb. Do you guys know why? Is this normal / expected? @syntonym and charles.

syntonym Over a year ago

What exactly did you do? If you converted to black and white you now have not a 3-tuple in each array cell but a single value - so the size consumption should be a third which seems to be the case for you. Seems like everything is normal.

Collectives™ on Stack Overflow

Python - Image from numpy array losing true pixels

1 Answer 1

4 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

4 Comments

Your Answer

Sign up or log in

Post as a guest

Related