The algorithm should known to public; but the key needs to be confidential.(Universal standard-Kerckhoffs's principle)

1. Key size

2. Performance/ Response time for Encryption or Decryption (depends on the system you use)

3. Mathematical proof for standardization of security provided by that algorithm

Key Size of course. 

But you wouldn't know the strength until it is broken. The breaking can be done in ways no one can define and you wouldn't know until you break it. But the strength of a key is measured in time in relation to brute force attack. The more time it takes more strong it is. All the keys can be broken but it is just a matter of time. For example, a 256 bits AES  key takes 3×10^51 years to break by brute force.

Read on Techniques of cryptanalysis

for signals processing

Imperceptibilty

Amplification

Noise addition

Low pass filtering

Requantization

Re-sampling

Compression

__________________________________

another aspects parameters

MSE

RMSE

UIQI

SNR

MAE

PSNR

AD

MD

NAE

Payload

In addition to the previous given parameters to check the strength of an information security algorithm, you can execute local Shannon entropy test, mean absolute error (MAE), contrast analysis, key sensitivity test, Chi-square test, etc.

For Images Encryption, the main parameters used for measurement are: NPCR, UACI, Entropy, and Correlation Coefficient for Columns and Rows, in addition to the key senstivity.

https://www.mathworks.com/matlabcentral/fileexchange/73037-matlab-aes-encryption-decryption-example

https://www.mathworks.com/matlabcentral/fileexchange/73037-matlab-aes-encryption-decryption-example

Most common algorithm for image cryptography: Chi-square test- Histogram- Correlation- Entropy- MSE- NPCR -PSNR-UACI-SSIM- Key sensitivity test.

least common algorithms: encryption and decryption time.

its resilience to different types of attacks, normally key size makes the difference in the success of attackers' cryptoanalysis trials.