Optimal use of BitmapFactory.Options.inSampleSize for speed

Question

Thanks to Schermvlieger for asking this question on anddev.org,

I'm just copying his question to SO as nobody replied on the other site and I'm also facing the same problem.

I was wondering what would be the optimal use of BitmapFactory.Options.inSampleSize with regards to speed of displaying the image.
The documentation mentions using values that are a power of 2, so I am working with 2, 4, 8, 16 etc.

The things I am wondering about are:

1. Should I resample down to the smallest size that is still larger than the screen resolution, or should I sample down to the size just enough to avoid an OutOfMemoryError?
2. How would one calculate the maximum size of an image that could still be displayed without running out of memory? Does the color-depth of the image play a role as well, and the depth of the display?
3. Is it efficient to display images via two mechanisms (BitmapFactory for large files, setImageURI() for smaller ones) I am using an ImageSwitcher by the way.
4. Would it help creating the Bitmap, BitmapFactory.Options and inTempStorage in the beginning of the application or creating them only on the fly, when needed?

Answer 1

You should always try to load and pre-scale images so that they are as close as possible to their final displayed size. Scaling images at drawing time is extremely expensive and should be avoided at all cost.

Considering the memory cost of an image, yes, the color-deptch plays a very important role. Images in ALPHA_8 format use 1 byte per pixel, images in RGB_565 or ARGB_4444 use 2 bytes per pixel and images in ARGB_8888 use 4 bytes per pixel. The depth of the display does not matter at all. You should always try to use ARGB_8888 to get the best possible quality, but 565 can be good enough if your image is opaque.