This behaviour depends on how the image file was found by the spider. If the image was part of the page (e.g. an image tag like <img src="pics/house.jpg">), then the search result will be a link to the page containing the image.
If the image was found via a link to the image file itself, (e.g. <a href="pics/house.jpg">this is a picture of my house</a>) then the search result will link to the image file itself.
If you always want the latter behaviour, perhaps you should index the images by themselves first, either via additional start points (making sure they are indexed before the pages), or by creating (or indexing a pre-existing) web page which linked directly to the images as in the above example.