Abstract: Atmel SAM3S + ESP8826 seems to be my best bet.
I searched around a bit more. As I already knew, if I were to use a STM or NXP mcu, it would cost 10+ dollars for a mcu with DCMI (digital camera interface). And I would still need a chip for the BLE radio.
Today I found out, that Atmel SAM S70 also supports a CMOS camera interface.
It looks neat, but also simple enough to implement with the logic blocks on the PSoC. At least the SAM S70 is cheaper (about 7 dollars(at 1k units)) than the NXP or ST chips, and also in a hobbyist friendlier package.
This webpage will also be very useful if I'll take the PSoC + camera interface with logic blocks approach. At 2.29 dollars (at 1k units) it would seem as the cheapest option, but that option comes without DMA or Universal Digital Blocks so it is useless for what I am doing. The ones that actually have the UDBs cost around 5 dollars.
And probably the best thing about it is that atmel has already written some examples with JPEG compression which is just fantastic.
Maybe I would use WiFI instead of BLE? It seems easer to implement but more expensive to build. Than again, an ESP8826 does not cost that much.